Machine Learning Integration in Cryptocurrency Trading: A Systematic Review of Fintech Implications

Machine Learning Integration in Cryptocurrency Trading: A Systematic Review of Fintech Implications | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Systematic Review Machine Learning Integration in Cryptocurrency Trading: A Systematic Review of Fintech Implications Péter Lengyel, János Pancsira, István Füzesi This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7411123/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 13 Feb, 2026 Read the published version in Discover Artificial Intelligence → Version 1 posted 13 You are reading this latest preprint version Abstract This review synthesizes research on fintech implications of integrating machine learning algorithms into cryptocurrency trading strategies to address the fragmented understanding of their impact on trading efficacy, risk management, and financial innovation. The review aimed to evaluate current knowledge on machine learning applications, benchmark algorithmic trading performance, identify risk mitigation techniques, compare algorithm effectiveness, and examine regulatory and ethical considerations. A systematic analysis of diverse methodologies, including supervised, reinforcement, and hybrid learning models across global computational finance and AI literature, was conducted. Findings indicate that deep learning and ensemble methods significantly enhance predictive accuracy and trading profitability under volatile market conditions, while reinforcement learning frameworks improve dynamic portfolio optimization and risk-adjusted returns. Risk management benefits arise from integrating technical indicators and reward-based safety mechanisms, though universal frameworks remain lacking. Fintech integration advances through blockchain-enabled transparency and automation, yet practical deployment faces scalability and interoperability challenges. Ethical and regulatory discourse is nascent, underscoring the need for responsible AI frameworks to ensure market integrity and investor protection. These findings collectively demonstrate that machine learning substantially transforms cryptocurrency trading strategies, offering enhanced performance and risk control within evolving fintech infrastructures, while highlighting critical gaps in regulatory compliance and ethical governance that warrant focused future research. Cryptocurrency trading Machine learning Reinforcement learning Fintech Risk management Systematic review Figures Figure 1 Figure 2 1. Introduction Since the emergence of Bitcoin in 2008, cryptocurrency markets have evolved into a global financial ecosystem, characterised by extreme volatility, decentralisation, and technological innovation (Nakamoto, 2008 ; Feng et al., 2019 ). These markets challenge traditional financial models due to their non-linear dynamics and speculative behaviour, making them an ideal environment for the application of machine learning (ML) techniques (Kristjanpoller & Minutolo, 2018 ). ML has increasingly been adopted in cryptocurrency trading for its ability to process vast amounts of data, identify patterns, and adapt to evolving market conditions. Techniques such as supervised learning, deep learning, and reinforcement learning (RL) offer promising alternatives to rule-based strategies (McNally et al., 2018 ; Alessandretti et al., 2018 ). Among these, Long Short-Term Memory (LSTM) networks and Gated Recurrent Units (GRU) have proven effective in time series forecasting, while ensemble models such as Random Forest and XGBoost demonstrate robustness against noise (Sebastião & Godinho, 2021 ; Lahmiri & Bekiros, 2019 ). Reinforcement learning approaches like Deep Q-Networks (DQN) and Proximal Policy Optimization (PPO) are particularly suited for trading applications due to their ability to model sequential decision-making and maximise cumulative returns (Yang et al., 2020 ). Furthermore, hybrid models combining sentiment analysis with price prediction have gained attention, leveraging data from social media platforms to enhance prediction accuracy (Mai et al., 2018 ). Despite these advances, several limitations remain. Overfitting, lack of interpretability, and insufficient generalisability are persistent technical concerns (Ji et al., 2019 ). Ethical and regulatory challenges, such as algorithmic bias, data privacy, and market manipulation, also pose significant barriers to widespread adoption (Tapscott & Tapscott, 2016 ; Zwitter et al., 2020 ). This study aims to systematically evaluate the integration of ML into cryptocurrency trading from a fintech perspective. By analysing 57 peer-reviewed studies, we assess the effectiveness of various ML models, examine risk mitigation practices, and identify both opportunities and challenges in practical implementation. The goal is to provide a structured understanding of how ML contributes to performance optimisation, ethical automation, and technological transformation in crypto-financial systems. 2. Methodology This study follows a systematic literature review approach to examine the application of machine learning (ML) in cryptocurrency trading and its implications for fintech. The methodology was designed to ensure comprehensive coverage, transparency, and replicability, following recognised best practices in systematic research. 2.1 Research Questions The review is structured around the following four research questions (RQs): RQ1 : How do ML algorithms enhance predictive accuracy and trading performance in cryptocurrency markets? RQ2 : What risk management techniques are facilitated by ML-driven trading strategies? RQ3 : How are these algorithms integrated into fintech infrastructures and automated trading systems? RQ4 : What ethical, regulatory, and practical limitations emerge in the use of ML for cryptocurrency trading? These questions guided both the search strategy and the data synthesis stages. 2.2 Literature Search and Study Selection A comprehensive search strategy was applied across major academic databases, including Scopus, Web of Science, IEEE Xplore, SpringerLink, and arXiv. The initial query, fintech implications of integrating ML in cryptocurrency trading , was expanded into focused expressions such as: Reinforcement learning for portfolio optimisation in crypto markets ML-based risk mitigation in crypto algorithmic trading Sentiment analysis and price prediction in blockchain assets The search covered peer-reviewed articles and high-quality preprints published between 2016 and 2025, limited to English-language studies directly addressing cryptocurrency trading. Inclusion criteria : Studies applying supervised, unsupervised, deep, or reinforcement learning to cryptocurrency markets; Papers reporting empirical results on prediction, trading performance, or risk control; Articles discussing fintech-relevant elements such as automation, blockchain, or regulation. Exclusion criteria : Studies on traditional finance only (e.g., stock or forex markets); Non-empirical works or those without reproducible methods; Duplicate or off-topic publications. The screening and selection process adhered to the PRISMA 2020 guidelines (Page et al., 2021 ), ensuring transparent and systematic identification, screening, and reporting. Of the initial set of 282 articles, 57 were retained after title/abstract screening and full-text evaluation. Citation chaining (both backward and forward) was used to supplement the final sample (see Fig. 1 ). The final set of articles was selected through a transparent screening pipeline using PRISMA principles, complemented by best practices in systematic data collection and visualisation (Bohr & Memarzadeh, 2020 ). Source: Own editing based on Page et al. ( 2021 ) 2.3 Data Extraction and Thematic Analysis Data were extracted along five key dimensions: Model characteristics : algorithm type (e.g., LSTM, GRU, DQN), learning framework, optimisation goal; Performance metrics : accuracy, F1 score, RMSE, Sharpe ratio, ROI; Risk management methods : Value at Risk (VaR), stop-loss strategies, reward shaping; Computational aspects : latency, backtesting, scalability; Fintech integration : blockchain infrastructure, DeFi mechanisms, smart contracts, and regulatory concerns. Thematic analysis was conducted using Braun and Clarke’s ( 2006 ) six-phase approach. This combined quantitative aggregation (e.g., number of studies using RL models) with qualitative coding (e.g., ethical implications, automation benefits). Each study was mapped to one or more of the research questions. Summarised findings are presented in the Results section, while detailed tabular comparisons are provided in Appendix A to maintain clarity and conciseness in the main body of the paper. 3. Results 3.1. Descriptive Overview of Reviewed ML-Based Crypto Trading Studies This section maps the research landscape of the literature on fintech implications of integrating machine learning algorithms into cryptocurrency trading strategies, encompassing a broad spectrum of methodologies, including supervised learning, reinforcement learning, deep learning, and hybrid approaches. The studies predominantly focus on predictive modeling for price forecasting, trading strategy optimization, and risk management within volatile cryptocurrency markets, often leveraging neural networks, decision trees, and reinforcement learning frameworks. Geographic and disciplinary diversity is evident, with contributions from computational finance, artificial intelligence, and ethical regulatory perspectives, reflecting the multifaceted nature of fintech innovation. This comparative analysis addresses key research questions by synthesizing empirical findings on algorithmic performance, risk mitigation, and the broader fintech ecosystem impacts, including ethical and regulatory considerations. To provide a comprehensive view of the literature, the following part of this subsection presents a thematic analysis structured around five key domains identified during the review: predictive accuracy, trading performance, risk management effectiveness, algorithmic efficiency, and fintech integration impact. This classification reflects the dominant methodological and conceptual trends observed across the selected articles. A detailed summary of all 57 reviewed studies, including key characteristics, applied machine learning techniques, and major findings, is presented in Appendix A (Appendix A. Summary Table of Reviewed Studies on ML in Cryptocurrency Trading). The appendix table is organised along the same five analytical dimensions, thus reinforcing the thematic synthesis and enhancing the transparency and reproducibility of this systematic review. Predictive Accuracy : Numerous studies (n = 40) reported that deep learning architectures such as LSTM and GRU offer substantial improvements in prediction accuracy over traditional models (Lu et al., 2024 ; Viéitez et al., 2024 ; Toleti et al., 2024 ). Ensemble methods and reinforcement learning strategies further enhance robustness across different market regimes (Huang, 2024 ; Jabbar & Jalil, 2024 ; Bouteska et al., 2024 ). Moreover, integrating sentiment analysis from social media and causal feature engineering boosts performance (Belcastro et al., 2023 ; Avramelou et al., 2023 ; Amirzadeh et al., 2023 ). Trading Performance : A total of 38 studies indicated that ML-based trading strategies outperformed traditional benchmarks like buy-and-hold, especially in terms of Sharpe ratio, return consistency, and drawdown reduction (Lua et al., 2025 ; Otabek & Choi, 2024 ; Asgari & Khasteh, 2021 ). Reinforcement learning models with novel reward structures yielded notable adaptability and profitability (Kang et al., 2024 ; Kouloumpris & Vlahavas, 2023 ; Lucarelli & Borrotti, 2019 ). Some researchers focused on portfolio optimisation and risk-adjusted return frameworks (Song, 2025 ; Ramkumar, 2021 ; Lucarelli et al., 2020). Risk Management Effectiveness : Thirty studies incorporated financial risk metrics—such as Value at Risk (VaR), drawdown controls, and the Sharpe Ratio, to evaluate and mitigate model-induced risks (Huang, 2024 ; Zhao et al., 2024 ; Ślepaczuk & Zenkova, 2019 ). Hybrid architectures using technical indicators in conjunction with ML methods enhanced the detection of market anomalies and reduced false trading signals (Tanwar & Raboaca, 2022 ; Bellocca et al., 2022 ). Reinforcement learning models featuring penalty mechanisms and conservative position sizing demonstrated prudent risk-reward trade-offs (Amirzadeh et al., 2023 ; Kumlungmak, 2024 ). Algorithmic Efficiency : Approximately 25 studies dealt with the practical implementation of ML models, focusing on scalability, real-time performance, and computational efficiency. Genetic algorithms and hyperparameter optimisation techniques were deployed to reduce overhead and improve performance (Lu et al., 2024 ; Su, 2024 ; Okasová & Košťál, 2024). Deep reinforcement learning with attention mechanisms proved effective in multi-asset scenarios and long sequence processing (Betancourt & Chen, 2021 ). Fintech Integration Impact : Around 20 studies tackled the broader fintech implications of ML integration. Key themes included regulatory compliance, explainability, fairness, and AI governance in crypto-finance ecosystems (Kumari et al., 2024 ; Pasupuleti, 2024 ; Alibašić, 2023 ). The convergence of AI and blockchain improved system transparency, fraud detection, and investor trust (Mandych et al., 2023 ; Hou, 2025 ). Despite these gains, the literature identified a need for consistent ethical and regulatory frameworks to ensure responsible deployment (Srivastava & Sikroria, 2024 ). 3.2. Temporal Trends in ML Applications for Cryptocurrency Trading The integration of machine learning into cryptocurrency trading strategies has evolved considerably from 2016 onwards, reflecting increasing sophistication in modeling techniques and expanding fintech implications. Early research primarily focused on applying basic machine learning models to price prediction and algorithmic trading in cryptocurrencies. Over time, studies incorporated deep learning and reinforcement learning to tackle market volatility and enhance decision-making in automated trading systems. Recent advancements emphasize hybrid models, sentiment analysis, ethical considerations, and the development of robust frameworks to optimize trading performance while addressing regulatory and operational challenges in the dynamic cryptocurrency market. A structured summary is provided in Table 2 . Table 2 Chronological Evolution of ML Applications in Cryptocurrency Trading (2016–2025) Years Research Direction Description 2016–2018 Early Machine Learning Applications in Cryptocurrency Trading Initial studies applied fundamental machine learning techniques such as support vector machines and neural networks for price prediction and trading signal generation. Research focused on exploiting market inefficiencies and developing algorithmic trading strategies that outperform traditional benchmarks, with preliminary insights into market sentiment and social media data integration. 2019–2020 Emergence of Deep Reinforcement Learning and Portfolio Optimization Research introduced deep reinforcement learning methods and advanced portfolio management frameworks to enhance profitability and risk-adjusted returns in cryptocurrency trading. Studies explored Q-learning variants and actor-critic models, integrating technical indicators and sentiment analysis to improve trading decisions and strategy robustness. 2021–2022 Advancement of Hybrid and Deep Learning Models for Automated Trading Focus shifted toward combining deep learning architectures like LSTM, CNN, and self-attention networks with reinforcement learning to develop adaptive and high-frequency trading algorithms. Emphasis was placed on improving predictive accuracy, managing market risks, and leveraging multi-factor models alongside social sentiment for enhanced trading performance. 2023 Integration of Causal Analysis, Ethical Frameworks, and Sentiment-Aware Models Research expanded to include causal analysis frameworks to empower reinforcement learning agents and the ethical implications of AI and ML use in cryptocurrency trading. Sentiment-aware deep reinforcement learning models and safety mechanisms were proposed to mitigate risk and improve decision-making under volatile market conditions. Ethical considerations and regulatory compliance became central themes alongside technical innovations. 2024 Optimization of Machine Learning Frameworks and Fintech Innovations Studies emphasized the optimization of genetic algorithms, hybrid models, and multi-agent reinforcement learning approaches to address market volatility and arbitrage opportunities. The integration of smart contracts and blockchain oracles with ML models facilitated transparent and automated portfolio management. Research also benchmarked different ML algorithms for their effectiveness in volatility prediction, risk adjustment, and profitability across diverse cryptocurrencies. 2025 Deployment of Advanced ML Systems in Real-World Trading Environments The latest research demonstrated that the deployment of ML-powered decentralized applications and trading bots achieves substantial returns on investment. Emphasis was placed on practical implementations of LSTM, gradient boosting, and reinforcement learning models within secure and trust-free fintech infrastructures, highlighting scalability, transparency, and adaptability in real-time cryptocurrency trading. Source Author’s own editing based on reviewed articles 3.3. Thematic Insights into ML Strategies and Fintech Integration The literature on integrating machine learning algorithms into cryptocurrency trading strategies reveals several dominant themes focused on enhancing trading performance, risk management, and fintech innovation. A large proportion of studies emphasize predictive modeling and algorithmic trading strategies employing diverse machine learning techniques such as deep learning, reinforcement learning, and ensemble methods. Significant attention is also given to managing the high volatility and risk inherent in cryptocurrency markets through advanced methods, including hybrid models and risk-adjusted portfolio optimization. Emerging research highlights ethical, regulatory, and operational challenges, underscoring the need for frameworks that ensure responsible AI use while fostering financial innovation in this rapidly evolving fintech landscape. The classification is summarised in Table 3 . Table 3 Dominant Research Themes in ML-based Cryptocurrency Trading (n = 57 papers) Theme Appears In Theme Description Machine Learning Models for Price Prediction and Trading Performance 43 Papers Extensive research focuses on employing machine learning models such as neural networks (LSTM, GRU), random forests, support vector machines, and ensemble methods to predict cryptocurrency prices and inform trading strategies. These models have demonstrated improved predictive accuracy and profitability over traditional benchmarks across various cryptocurrencies under diverse market conditions (Lu et al., 2024 ; Huang, 2024 ; Viéitez et al., 2024 ; Song, 2025 ; Jabbar & Jalil, 2024 ; Hafid et al., 2024 ; Park & Seo, 2022 ; Bouteska et al., 2024 ; Sebastião & Godinho, 2021 ; Singh et al., 2022 ; Li et al., 2024 ). Reinforcement Learning for Algorithmic Trading and Portfolio Optimization 24 Papers Reinforcement learning (RL) techniques, including Deep Q-networks, Advantage Actor-Critic, Proximal Policy Optimization, and multi-agent frameworks, are widely applied to develop adaptive trading strategies that optimize asset allocation and manage risk. Studies show RL's capability to handle dynamic market environments and improve returns while balancing risk, especially in volatile cryptocurrency markets (Kang et al., 2024 ; Otabek & Choi, 2024 ; Kouloumpris & Vlahavas, 2023 ; Amirzadeh et al., 2023 ; Zhao et al., 2024 ; Peng et al., 2022 ; Asgari & Khasteh, 2021 ; Felizardo et al., 2022 ; Betancourt & Chen, 2021 ; Lucarelli et al., 2020; Bu & Cho, 2018 ; Lucarelli & Borrotti, 2019 ; Sattarov et al., 2020 ; Kumlungmak, 2024 ). Risk Management and Volatility Mitigation Using Machine Learning 23 Papers Addressing cryptocurrency market volatility and associated risks is a key focus, with machine learning enabling advanced risk mitigation techniques such as EGARCH integration, stochastic neural networks, hybrid models, and portfolio optimization frameworks that incorporate risk-adjusted metrics like Sharpe ratio and Value at Risk. These approaches enhance trading robustness and stability (Lu et al., 2024 ; Song, 2025 ; Kouloumpris & Vlahavas, 2023 ; Zhao et al., 2024 ; Tanwar & Raboaca, 2022 ; Alaminos et al., 2024 ; Chen, 2022 ; Koker & Koutmos, 2020 ). Integration of Sentiment Analysis and External Data Sources 14 Papers Several studies incorporate sentiment analysis from social media and market indicators to enhance price prediction models. The integration of natural language processing and sentiment scoring improves understanding of market dynamics and contributes to more informed trading decisions, although the impact varies across models and cryptocurrencies (Viéitez et al., 2024 ; Belcastro et al., 2023 ; Avramelou et al., 2023 ; Leung et al., 2023 ; Pang et al., 2019 ). Ethical, Regulatory, and Operational Implications of AI in Cryptocurrency Trading 9 Papers Emerging literature explores ethical frameworks and regulatory challenges posed by AI and machine learning in cryptocurrency trading, emphasizing the need for responsible AI deployment, transparency, market integrity, investor protection, and novel regulatory approaches tailored to digital assets’ unique characteristics (Kumari et al., 2024 ; Srivastava & Sikroria, 2024 ; Pasupuleti, 2024 ; Mandych et al., 2023 ; Alibašić, 2023 ). Hybrid and Ensemble Methodologies for Enhanced Trading Strategies 12 Papers Hybrid models combining deep learning architectures (e.g., CNN-LSTM, LSTM-GRU) and ensemble learning methods have been proposed to leverage complementary strengths of different algorithms, improving predictive performance and trading profitability in volatile cryptocurrency markets (Park & Seo, 2022 ; Bouteska et al., 2024 ; Bellocca et al., 2022 ; Qing et al., 2022a ). Automated Trading Systems and Decentralized Finance Integration 8 Papers Research includes blockchain and smart contract integration with machine learning-based trading systems, enhancing transparency, automation, and security in cryptocurrency trading. These systems reduce barriers for investors and enable decentralized portfolio management (Lua et al., 2025 ; Mandych et al., 2023 ; Hou, 2025 ). Arbitrage Detection and Statistical Trading Techniques Using Machine Learning 4 Papers Machine learning models are utilized to identify arbitrage opportunities and apply statistical arbitrage trading strategies in cryptocurrency markets, offering potential for riskless profits but facing challenges due to rapid market changes (Okasová & Košťál, 2024; Fischer et al., 2019 ). Source Author’s own editing based on reviewed articles 3.4. Cross-Study Evaluation of Methodologies and Fintech Implications The reviewed literature on the fintech implications of integrating ML algorithms into cryptocurrency trading strategies reveals a dynamic and rapidly evolving field marked by innovative methodologies and promising empirical results. A key theme is the demonstrated potential of ML models to enhance predictive accuracy, trading profitability, and risk management in highly volatile cryptocurrency markets. However, the diversity of approaches and varying data quality pose challenges to generalizability and robustness. Additionally, ethical, regulatory, and operational considerations remain underexplored, highlighting critical gaps for future research. This synthesis critically evaluates the strengths and limitations across methodological rigor, algorithmic performance, risk mitigation, fintech integration, and ethical/regulatory discourse. The analysis is based on the full set of 57 reviewed studies and is summarised in a comparative matrix provided in Table 4 . The table includes study-specific evaluations aligned with each analytical category and highlights both strengths and limitations. This structured synthesis aims to guide future research by clarifying which areas exhibit robust evidence and where significant knowledge gaps remain. Table 4 Comparative Matrix of Strengths and Weaknesses in Reviewed Studies on Machine Learning for Cryptocurrency Trading Aspect Strengths Weaknesses Methodological Approaches The literature employs a broad spectrum of ML techniques, including neural networks, reinforcement learning, genetic algorithms, and ensemble methods, demonstrating adaptability to complex market dynamics and non-linear patterns (Lu et al., 2024 ; Su, 2024 ; Song, 2025 ; Otabek & Choi, 2024 ). Rigorous backtesting, cross-validation, and real-world trading simulations enhance the credibility of findings (Huang, 2024 ; Jabbar & Jalil, 2024 ; Sattarov et al., 2020 ). The integration of hybrid models, such as combining deep learning with Hidden Markov Models or causal analysis, further strengthens predictive capabilities (Toleti et al., 2024 ; Amirzadeh et al., 2023 ). Despite methodological diversity, many studies rely heavily on historical data with limited temporal scope, which may not capture evolving market regimes or rare events (Viéitez et al., 2024 ; Belcastro et al., 2023 ). Some models exhibit overfitting risks due to high parameterization and insufficient out-of-sample validation (Srivastava & Sikroria, 2024 ; Sebastião & Godinho, 2021 ). The heterogeneity in data preprocessing and feature selection complicates direct comparison and replication (Sruthi & Shahithabanu, 2024 ; Sebastião & Godinho, 2021 ). Moreover, few studies address the impact of transaction costs and slippage comprehensively (Lam & Makarov, 2023 ). Algorithmic Performance and Predictive Accuracy ML models, particularly LSTM, GRU, Random Forest, and reinforcement learning algorithms like PPO and DDPG, consistently outperform traditional benchmarks such as buy-and-hold strategies, achieving significant returns and improved Sharpe ratios (Lu et al., 2024 ; Lua et al., 2025 ; Kang et al., 2024 ; Otabek & Choi, 2024 ; Lucarelli & Borrotti, 2019 ). The use of sentiment analysis and social media data enhances prediction robustness in some cases (Viéitez et al., 2024 ; Belcastro et al., 2023 ; Avramelou et al., 2023 ). Reinforcement learning frameworks demonstrate adaptability to market volatility and dynamic portfolio optimization (Kang et al., 2024 ; Zhao et al., 2024 ; Kumlungmak, 2024 ). Performance gains are often context-specific, with some models showing diminished effectiveness during extreme market conditions or bearish phases (Sebastião & Godinho, 2021 ; Kumlungmak, 2024 ). Sentiment analysis integration yields mixed results, with limited incremental benefit in certain studies (Viéitez et al., 2024 ). The computational complexity and training time of deep models may hinder real-time applicability (Huang, 2024 ; Lam & Makarov, 2023 ). Additionally, the lack of standardized evaluation metrics and inconsistent reporting of risk-adjusted returns limit comprehensive assessment (Jabbar & Jalil, 2024 ; Srivastava & Sikroria, 2024 ). Risk Management and Volatility Handling Several studies incorporate risk-adjusted metrics such as Sharpe ratio, Value at Risk (VaR), and drawdown controls, demonstrating ML’s capacity to balance profitability with risk mitigation (Huang, 2024 ; Kang et al., 2024 ; Zhao et al., 2024 ; Koker & Koutmos, 2020 ). Hybrid approaches combining technical indicators with ML models improve robustness against market noise and volatility (Lu et al., 2024 ) (Kouloumpris & Vlahavas, 2023 ) (Alaminos et al., 2024 ). Reinforcement learning agents are designed with reward functions and safety mechanisms to reduce exposure during uncertain periods (Kouloumpris & Vlahavas, 2023 ). Risk management strategies are often heuristic or model-specific, lacking universal frameworks applicable across different cryptocurrencies or market conditions (Kouloumpris & Vlahavas, 2023 ; Kumlungmak, 2024 ). Some reinforcement learning models prioritize profit maximization at the expense of risk awareness, leading to potential overexposure (Kouloumpris & Vlahavas, 2023 ; Asgari & Khasteh, 2021 ). The challenge of modeling extreme tail risks and sudden market shocks remains insufficiently addressed (Alaminos et al., 2024 ). Furthermore, the impact of leverage and liquidity constraints is rarely considered (Fischer et al., 2019 ). Integration within Fintech Infrastructure The integration of ML models with blockchain technologies, smart contracts, and decentralized applications (dApps) enhances transparency, automation, and accessibility in cryptocurrency trading (Lua et al., 2025 ) (Pasupuleti, 2024 ; Mandych et al., 2023 ). Studies highlight the potential for ML-driven portfolio management systems to reduce barriers for retail investors and improve operational efficiency (Lua et al., 2025 ; Ramkumar, 2021 ). The use of oracles and on-chain verification mechanisms supports real-time, trustless trading environments (Lua et al., 2025 ). Practical deployment challenges include scalability, latency, and security vulnerabilities inherent in blockchain-based systems (Mandych et al., 2023 ). The complexity of integrating ML predictions with decentralized finance (DeFi) protocols is underexplored (Pasupuleti, 2024 ). Many studies remain theoretical or simulation-based without full implementation in live fintech ecosystems (Sruthi & Shahithabanu, 2024 ; Hou, 2025 ). Additionally, interoperability issues between ML models and existing financial infrastructures pose barriers (Pasupuleti, 2024 ). Ethical, Regulatory, and Operational Considerations Emerging research acknowledges the ethical implications of AI and ML in cryptocurrency trading, emphasizing the need for responsible frameworks to ensure market integrity and investor protection (Alibašić, 2023 ). Discussions include the risks of algorithmic biases, market manipulation, and the necessity for adaptive regulatory approaches tailored to digital assets (Srivastava & Sikroria, 2024 ; Mandych et al., 2023 ; Alibašić, 2023 ). The potential for AI to enhance compliance through RegTech applications is recognized (Pasupuleti, 2024 ). Ethical and regulatory aspects are insufficiently addressed in most empirical studies, with limited operational guidelines or frameworks proposed (Sruthi & Shahithabanu, 2024 ; Alibašić, 2023 ). The fast-paced evolution of ML-driven trading outstrips current regulatory capacities, raising concerns about transparency and accountability (Srivastava & Sikroria, 2024 ; Mandych et al., 2023 ). Operational challenges such as model interpretability, data privacy, and systemic risks remain largely unexplored (Alibašić, 2023 ). The literature lacks consensus on best practices for balancing innovation with ethical constraints (Alibašić, 2023 ). Data Quality and Feature Engineering Studies leverage diverse data sources, including technical indicators, market prices, social media sentiment, and blockchain metrics, enriching model inputs and enhancing predictive power (Viéitez et al., 2024 ; Belcastro et al., 2023 ; Avramelou et al., 2023 ; Li et al., 2024 ). Advanced feature selection and causal analysis improve model interpretability and relevance (Amirzadeh et al., 2023 ; Li et al., 2024 ). Data limitations include short historical records, especially for newer cryptocurrencies, leading to potential biases and reduced model generalizability (Park & Seo, 2022 ; Alaminos et al., 2024 ). The noisy and unstructured nature of social media data challenges sentiment analysis accuracy (Viéitez et al., 2024 ; Leung et al., 2023 ). Inconsistent data preprocessing and lack of standardized datasets hinder reproducibility and benchmarking (Sruthi & Shahithabanu, 2024 ; Avramelou et al., 2023 ). The impact of data latency and real-time availability on model performance is rarely discussed (Lua et al., 2025 ). Comparative Evaluation of ML Algorithms Comprehensive comparisons reveal that ensemble methods and deep learning architectures often outperform simpler models, with tree-based models like Random Forest excelling in return predictability (Jabbar & Jalil, 2024 ) (Bouteska et al., 2024 ) (Li et al., 2024 ). Reinforcement learning approaches show promise in dynamic strategy adaptation and portfolio management (Kang et al., 2024 ; Otabek & Choi, 2024 ; Kumlungmak, 2024 ). Hybrid models combining multiple ML techniques achieve superior results (Toleti et al., 2024 ; Amirzadeh et al., 2023 ). Comparative studies are limited by inconsistent experimental setups, varying datasets, and differing evaluation metrics, complicating definitive conclusions (Jabbar & Jalil, 2024 ) (Bouteska et al., 2024 ). Some algorithms require extensive hyperparameter tuning and computational resources, limiting practical deployment (Huang, 2024 ; Srivastava & Sikroria, 2024 ). The trade-offs between model complexity, interpretability, and real-time applicability are insufficiently explored (Felizardo et al., 2022 ; Lam & Makarov, 2023 ). Few studies address the robustness of algorithms across diverse market regimes (Sebastião & Godinho, 2021 ). Source Author’s own editing based on reviewed articles To enhance the clarity of the comparative findings, a visual summary is provided in Fig. 2 . The heatmap highlights the relative strengths and limitations of key machine learning models based on six evaluation criteria: predictive accuracy, return on investment, scalability, risk management capability, interpretability, and regulatory alignment. The numerical ratings (on a 1–5 scale) are derived from the authors’ synthesis of the reviewed literature and represent a qualitative aggregation of model characteristics discussed in Sections 3.1 to 3.4 . This visualisation complements Table 4 and supports readers in quickly identifying which models are most suitable under specific constraints and objectives. Source: Authors’ compilation based on reviewed articles As shown in Fig. 2 , deep learning models such as LSTM and GRU demonstrate high predictive accuracy and solid ROI, but they generally lack interpretability and scalability. In contrast, traditional ensemble methods like Random Forest and XGBoost offer better scalability and interpretability, though at the cost of slightly lower predictive performance. Reinforcement learning (RL) techniques appear particularly effective in risk-sensitive environments, achieving high ROI and strong risk management capabilities, albeit with limitations in real-time scalability. Hybrid and ensemble models generally offer a balanced trade-off across most dimensions, but their complexity may hinder deployment in resource-constrained environments. These distinctions underscore the importance of aligning model choice with specific application goals, for example, using RL models in highly dynamic and volatile markets, whereas Random Forest may be more appropriate when interpretability and compliance requirements are prioritised. 3.5. Theoretical and Practical Implications To bridge the gap between academic inquiry and real-world application, this section synthesises the theoretical contributions and practical implications of integrating machine learning (ML) into cryptocurrency trading. The reviewed literature indicates how advanced computational models not only reshape predictive finance theory but also provide actionable insights for traders, fintech developers, and policymakers. Theoretical Implications The integration of machine learning algorithms into cryptocurrency trading strategies substantiates the hypothesis that ML can significantly enhance predictive accuracy and trading profitability. These findings support the view that advanced computational models outperform traditional financial models in volatile markets, thus challenging classical efficient market hypotheses by revealing exploitable inefficiencies in cryptocurrency markets (Alessandretti et al., 2018 ; Jabbar & Jalil, 2024 ; Lu et al., 2024 ). Reinforcement learning (RL) and deep learning models, particularly those employing hybrid and multi-agent frameworks, contribute to theoretical developments by addressing dynamic market conditions and risk management. Their adaptability and robustness in non-stationary environments expand the theory of algorithmic trading under uncertainty (Asgari & Khasteh, 2021 ; Kang et al., 2024 ; Kumlungmak, 2024 ). The incorporation of sentiment analysis and social media data into ML-based forecasting introduces a behavioural dimension to price prediction theories. This highlights the influence of investor sentiment and informational flows on cryptocurrency price dynamics, bridging finance, behavioural economics, and data science (Avramelou et al., 2023 ; Belcastro et al., 2023 ; Pang et al., 2019 ). Ethical and regulatory aspects arising from AI-driven trading challenge conventional frameworks of market efficiency and fairness. These findings suggest the need for expanded models that integrate the normative and institutional consequences of algorithmic decision-making in financial markets (Alibašić, 2023 ; Mandych et al., 2023 ; Srivastava & Sikroria, 2024 ). Comparative evaluations of ML algorithms indicate that ensemble and deep learning approaches, such as gradient boosting and LSTM networks, excel at capturing short- and long-term temporal dependencies in highly volatile price series. These insights refine theoretical models of time series forecasting and financial prediction (Alessandretti et al., 2018 ; Bouteska et al., 2024 ; Li et al., 2024 ). Practical Implications The consistent outperformance and robustness of ML-based trading strategies suggest considerable potential for fintech firms and individual traders to optimise portfolio management and risk mitigation. The findings support the broader application of AI-powered tools for real-time decision-making in digital asset markets (Jabbar & Jalil, 2024 ; Lua et al., 2025 ; Lucarelli & Borrotti, 2019 ). Seamless integration of ML models with blockchain infrastructure—such as smart contracts and oracles—offers practical pathways to transparent, decentralised, and automated trading systems. These innovations lower participation barriers and contribute to financial inclusion (Lua et al., 2025 ; Mandych et al., 2023 ; Pasupuleti, 2024 ). The identified ethical and regulatory challenges call for the development of governance frameworks and compliance mechanisms specifically tailored to AI-driven cryptocurrency trading. Addressing algorithmic biases, fraud prevention, and regulatory transparency is vital for ensuring sustainable fintech innovation (Alibašić, 2023 ; Srivastava & Sikroria, 2024 ). Reinforcement learning and multi-agent systems demonstrate practical utility in dynamic asset allocation and volatility management. Their ability to adapt to changing market regimes presents a valuable toolset for building resilient trading systems (Kang et al., 2024 ; Kumlungmak, 2024 ; Zhao et al., 2024 ). The use of sentiment analysis and alternative data sources in ML pipelines significantly enhances predictive capabilities. These findings suggest that fintech applications should integrate multimodal data to improve market forecasts and trading signal reliability (Avramelou et al., 2023 ; Belcastro et al., 2023 ; Leung et al., 2023 ). Despite the promising outcomes, issues such as limited model generalisability, sensitivity to extreme market shocks, and unaccounted transaction costs remain critical concerns. This underscores the need for continuous empirical validation and iterative model development to ensure scalability and operational viability in real-world trading environments (Alessandretti et al., 2018 ; Huang, 2024 ; Jabbar & Jalil, 2024 ). 4. Discussion This section critically interprets the reviewed literature in relation to the study’s core research questions and thematic domains. The aim is to provide a structured reflection on how the integration of machine learning (ML) into cryptocurrency trading strategies has influenced predictive performance, risk mitigation, fintech integration, and regulatory preparedness. 4.1 Predictive Modelling and Trading Performance The literature strongly supports the hypothesis that ML models outperform traditional statistical methods in cryptocurrency price forecasting. In particular, long short-term memory (LSTM), gated recurrent units (GRU), random forests, and ensemble methods show consistent improvements in return predictability. These results are robust across different cryptocurrencies, time frames, and feature sets. However, their effectiveness is often context-dependent, with some models underperforming in bearish or high-volatility conditions. This indicates that while ML enhances predictive power, its generalisation remains limited by data quality and market dynamics. Nevertheless, the findings confirm that ML can significantly improve trading profitability, partially answering RQ1 regarding the impact on trading outcomes. 4.2 Algorithmic Robustness and Adaptability Reinforcement learning (RL) algorithms, especially those using Proximal Policy Optimisation (PPO) and Deep Q-Learning, demonstrate the ability to adapt dynamically to shifting market conditions. Multi-agent RL systems allow decentralised decision-making and portfolio optimisation in real-time environments. However, these systems often require substantial computational resources and tuning. Despite their complexity, RL approaches provide flexibility and resilience, particularly in algorithmic trading, supporting RQ2 by showing how ML can optimise decision-making under uncertainty. 4.3 Risk Management and Volatility Handling Many studies incorporate risk-adjusted metrics such as the Sharpe ratio, drawdown, and Value at Risk (VaR), indicating an evolving maturity in modelling both return and risk. Hybrid models, which combine technical indicators and sentiment analysis, provide greater stability in volatile environments. Yet, extreme tail risks, leverage effects, and slippage remain underexplored. This suggests that ML models still lack comprehensive frameworks for high-stress scenarios. The research partially answers RQ3, demonstrating that ML improves volatility handling, but further development is needed for robust risk management across diverse market regimes. 4.4 Fintech Integration and Technical Feasibility The integration of ML with blockchain technologies (e.g., smart contracts, decentralised applications) reveals innovative approaches for automated and transparent trading systems. Oracles and trustless verification mechanisms further support scalability. However, interoperability, latency, and cybersecurity risks hinder real-world deployment. Many proposals remain theoretical or simulation-based. Thus, while ML shows promise in fintech innovation, RQ4 remains only partially answered, ML integration is feasible, but operational barriers constrain widespread implementation. 4.5 Ethical, Regulatory, and Operational Considerations The literature reveals growing awareness of ethical implications, including algorithmic bias, market manipulation, and investor protection. Discussions on RegTech applications and adaptive regulation frameworks highlight a path forward for responsible AI use. Nevertheless, few empirical studies offer actionable governance models, and most ignore legal compliance, data privacy, and explainability. This shows that RQ5 remains unanswered mainly, although ethical discourse is emerging, operational guidelines and regulatory alignment are still nascent. 5. Conclusion This systematic review examined the fintech implications of integrating machine learning (ML) algorithms into cryptocurrency trading strategies, analysing 57 peer-reviewed studies across five key thematic domains: predictive performance, algorithmic trading optimisation, risk management, technological integration, and ethical-regulatory concerns. Regarding RQ1 (How does ML affect trading performance in cryptocurrency markets?), the evidence overwhelmingly supports the notion that ML algorithms, especially deep learning models such as LSTM and GRU, significantly enhance predictive accuracy and trading returns (Lu et al., 2024 ; Jabbar & Jalil, 2024 ; Viéitez et al., 2024 ). ML-based strategies consistently outperform traditional methods across various coins and timeframes, especially under favourable market regimes. In relation to RQ2 (How effective are reinforcement learning and hybrid models for portfolio optimisation?), the review found strong support for the adaptability of reinforcement learning (RL) frameworks, particularly in volatile markets. Advanced RL methods and hybrid models using actor-critic structures and multi-agent systems have demonstrated superior risk-return profiles and dynamic rebalancing capabilities (Kang et al., 2024 ; Asgari & Khasteh, 2021 ). For RQ3 (To what extent can ML models manage volatility and mitigate trading risk?), most studies showed improved handling of short-term volatility and false trading signals through hybrid techniques and risk-adjusted metrics like the Sharpe Ratio and Value at Risk (Zhao et al., 2024 ; Tanwar & Raboaca, 2022 ). However, comprehensive tail-risk modelling and robustness under black-swan events remain underexplored. RQ4 (What is the feasibility of integrating ML models within real-world fintech infrastructure?) yielded mixed results. Although many studies propose ML-enhanced trading systems using blockchain, smart contracts, and decentralised automation (Lua et al., 2025 ; Mandych et al., 2023 ), real-world deployment is rare due to scalability issues, infrastructure latency, and interoperability concerns. Finally, RQ5 (How are ethical, regulatory, and governance issues addressed in ML-driven crypto trading?) remains insufficiently answered. While some recent works acknowledge the need for explainability, fairness, and investor protection (Kumari et al., 2024 ; Srivastava & Sikroria, 2024 ), concrete legal and ethical frameworks for responsible AI deployment in crypto trading are still lacking. In conclusion, while the integration of ML into cryptocurrency trading demonstrates substantial advancements in predictive analytics, profitability, and risk control, practical and normative challenges persist. Future research should prioritise the development of robust, interpretable, and regulation-aligned ML systems to ensure both market efficiency and investor safety in the rapidly evolving fintech ecosystem. 6. Limitations and Future Research Although this review provides a broad overview of the literature on machine learning applications in cryptocurrency trading, several limitations should be acknowledged. The study is restricted to peer-reviewed articles published in English between 2016 and 2025, which may have excluded relevant insights from grey literature, industry reports, or non-English academic sources. This language and publication bias may limit the global generalisability of the findings, particularly in regions where fintech innovations emerge outside the traditional academic ecosystem. The thematic scope of the review focused primarily on studies integrating machine learning algorithms into technical trading frameworks. As a result, other relevant perspectives—such as behavioural finance, institutional economics, or legal analyses of algorithmic trading—received limited attention. Moreover, the majority of the reviewed studies relied on historical backtesting or simulated trading environments. While these approaches offer valuable experimental insights, they do not necessarily reflect the constraints and uncertainties of real-world implementation, such as latency, transaction costs, or changing regulatory environments. Another concern is the potential for publication bias, as positive or novel outcomes are more likely to be published than negative or inconclusive findings. This may distort the perceived effectiveness of certain algorithms or modelling strategies, leading to overly optimistic interpretations of their practical viability. Looking ahead, future research should prioritise empirical studies that validate machine learning models under live market conditions and quantify the impact of operational constraints on trading performance. A promising direction is the integration of explainable AI techniques and ethical design principles into cryptocurrency trading algorithms to enhance transparency, accountability, and regulatory compliance. Further exploration is also needed in merging decentralised finance infrastructures with machine learning systems, including smart contracts, blockchain oracles, and automated market-making mechanisms. Another area ripe for investigation involves enhancing model robustness under extreme market events and high-volatility scenarios. Stress-testing ML strategies during black-swan events or regulatory interventions could provide deeper insights into their reliability. Finally, as data sources become increasingly diverse and unstructured, developing models capable of fusing multimodal inputs, including on-chain data, macroeconomic indicators, and real-time sentiment, will be crucial for generating actionable trading signals. By addressing these limitations and expanding the research horizon, scholars and practitioners can better navigate the complex interplay between artificial intelligence and digital financial markets. Declarations FUNDING: This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors. COMPETING INTERESTS: The authors declare no competing interests. AUTHOR CONTRIBUTIONS: P.L. conceived the study, developed the methodological framework, coordinated the research process, and contributed to the manuscript writing. J.P. performed the literature review, data analysis, and prepared the comparative tables. I.F. contributed to the interpretation of results and manuscript revision. All authors reviewed and approved the final manuscript. DATA AVAILABILITY: The study is based on publicly available academic literature and does not rely on proprietary or primary datasets. A full list of reviewed studies is included in Appendix A. Any additional materials or coding frameworks used in the thematic synthesis are available from the authors upon reasonable request. ETHICAL APPROVAL: This article does not contain any studies with human participants performed by any of the authors. INFORMED CONSENT: This article does not contain any studies with human participants performed by any of the authors. CONSENT TO PARTICIPATE: Not applicable. CONSENT TO PUBLISH: The authors consent to the publication of this manuscript. CLINICAL TRIAL NUMBER: Not applicable. References Alaminos D, Salas MB, Callejón-Gil ÁM. Managing extreme cryptocurrency volatility in algorithmic trading: EGARCH via genetic algorithms and neural networks. Quant Finance Econ. 2024;8(1):153–209. https://doi.org/10.3934/qfe.2024007 . Alessandretti L, ElBahrawy A, Aiello LM, Baronchelli A. (2018). Anticipating cryptocurrency prices using machine learning. Complexity . 2018(1), 16p. https://doi.org/10.1155/2018/8983590 Alibašić H. (2023). Developing an Ethical Framework for Responsible Artificial Intelligence (AI) and Machine Learning (ML) Applications in Cryptocurrency Trading: A Consequentialism Ethics Analysis. FinTech . 2(3), pp. 430–443. https://doi.org/10.3390/fintech2030024 Amirzadeh R, Nazari A, Thiruvady DR. Applying Artificial Intelligence in Cryptocurrency Markets: A Survey. Algorithms. 2022;15(11):428. https://doi.org/10.3390/a15110428 . Amirzadeh R, Thiruvady DR, Nazari A, Ee MS. (2023). A Framework for Empowering Reinforcement Learning Agents with Causal Analysis: Enhancing Automated Cryptocurrency Trading. arXiv.Org . https://doi.org/10.48550/arxiv.2310.09462 Asgari M, Khasteh H. (2021). Profitable Strategy Design for Trades on Cryptocurrency Markets with Machine Learning Techniques. arXiv: Trading and Market Microstructure . https://doi.org/10.48550/arxiv.2105.06827 Avramelou L, Nousi P, Passalis N, Doropoulos S, Tefas A. (2023). Cryptosentiment: A Dataset and Baseline for Sentiment-Aware Deep Reinforcement Learning for Financial Trading. 2023 IEEE International Conference on Acoustics, Speech, and Signal Processing Workshops (ICASSPW) , Rhodes Island, Greece, 2023, 1–5. https://doi.org/10.1109/icasspw59220.2023.10193330 Belcastro L, Carbone D, Cosentino C, Marozzo F, Trunfio P. Enhancing Cryptocurrency Price Forecasting by Integrating Machine Learning with Social Media and Market Data. Algorithms. 2023;16(12):542. https://doi.org/10.3390/a16120542 . Bellocca GP, Attanasio G, Cagliero L, Fior J. Leveraging the momentum effect in machine learning-based cryptocurrency trading. Mach Learn Appl. 2022;8:100310. https://doi.org/10.1016/j.mlwa.2022.100310 . Betancourt C, Chen WH. Reinforcement Learning with Self-Attention Networks for Cryptocurrency Trading. Appl Sci. 2021;11(16):7377. https://doi.org/10.3390/APP11167377 . Bohr A, Memarzadeh K. (2020). Chapter 12 - Systematic Review and Meta-Analysis. In A. Bohr & K. Memarzadeh, editors, Artificial Intelligence in Healthcare , 267–285. Academic Press. https://doi.org/10.1016/B978-0-12-818438-7.00012-5 Bouteska A, Abedin MZ, Hájek P, Yuan K. Cryptocurrency price forecasting – A comparative analysis of ensemble learning and deep learning methods. Int Rev Financial Anal. 2024;92. https://doi.org/10.1016/j.irfa.2023.103055 . Braun V, Clarke V. Using thematic analysis in psychology. Qualitative Res Psychol. 2006;3(2):77–101. https://doi.org/10.1191/1478088706qp063oa . Bu SJ, Cho S-B. Learning Optimal Q-Function Using Deep Boltzmann Machine for Reliable Trading of Cryptocurrency. In: Yin H, Camacho D, Novais P, Tallón-Ballesteros A, editors. Intelligent Data Engineering and Automated Learning – IDEAL 2018. IDEAL 2018. Lecture Notes in Computer Science(). Volume 11314. Cham: Springer; 2018. https://doi.org/10.1007/978-3-030-03493-1_49 . Chen Z. (2022). A Novel Machine Learning-assisted Pairs Trading Approach for Trading Risk Reduction. 2022 IEEE 1st Global Emerging Technology Blockchain Forum: Blockchain & Beyond (iGETblockchain ), Irvine, CA, USA, 2022, 1–6 https://doi.org/10.1109/iGETblockchain56591.2022.10087166 Felizardo L, Paiva FCL, Graves CdeV, Matsumoto EY, Costa AHR, Del-Moral-Hernandez E, Brandimarte P. Outperforming algorithmic trading reinforcement learning systems: A supervised approach to the cryptocurrency market. Expert Syst Appl. 2022;202. https://doi.org/10.1016/j.eswa.2022.117259 . Feng W, Wang Y, Zhang Z. Can cryptocurrencies be a safe haven: a tail risk perspective analysis. Appl Econ. 2019;51(47):5091–109. https://doi.org/10.1080/00036846.2019.1606405 . Fischer TG, Krauss C, Deinert A. Statistical Arbitrage in Cryptocurrency Markets. J Risk Financial Manag. 2019;12(1):31. https://doi.org/10.3390/JRFM12010031 . Hafid A, Rahouti M, Kong L, Ebrahim M, Serhani MA. Predicting Bitcoin Market Trends with Enhanced Technical Indicator Integration and Classification Models. arXiv: Mach Learn. 2024. https://doi.org/10.48550/arxiv.2410.06935 . Hou C. AI Technology’s Application and Impact in the Secondary Market of Virtual Currencies. J Appl Econ Policy Stud. 2025;16. https://doi.org/10.54254/2977-5701/2025.20565 . Huang S. Enhancing investment strategies through machine learning: A comprehensive analysis across market sectors. Appl Comput Eng. 2024;104. https://doi.org/10.54254/2755-2721/104/20240909 . Jabbar A, Jalil SQ. (2024). A Comprehensive Analysis of Machine Learning Models for Algorithmic Trading of Bitcoin. arXiv: Trading and Market Microstructure . https://doi.org/10.48550/arxiv.2407.18334 Ji Q, Zhang D, Zhao Y. Searching for safe-haven assets during the COVID-19 pandemic. Int Rev Financial Anal. 2019;71:101526. https://doi.org/10.1016/j.irfa.2020.101526 . Kang C, Woo J-S, Hong JW. (2024). A2C Reinforcement Learning for Cryptocurrency Trading and Asset Management. 2024 IEEE International Conference on Blockchain and Cryptocurrency (ICBC) . https://doi.org/10.1109/icbc59979.2024.10634413 Koker TE, Koutmos D. Cryptocurrency Trading Using Machine Learning. J Risk Financial Manag. 2020;13(8). https://doi.org/10.3390/JRFM13080178 . Kouloumpris E, Vlahavas I. Combining deep reinforcement learning with technical analysis and trend monitoring on cryptocurrency markets. Neural Comput Appl. 2023;35:21445–62. https://doi.org/10.1007/s00521-023-08516-x . Kristjanpoller W, Minutolo M. (2018). Forecasting volatility of cryptocurrencies using GARCH models. Applied Sciences , 8(10), 1987. https://doi.org/10.3390/app8101987 Kumari M, Sinha R, Chakrabarty P, Bhardwaj YK, Priya SA, Gupta S. (2024). Enhancing Digital Currency Pricing with Machine Learning Models. 2024 15th International Conference on Computing Communication and Networking Technologies (ICCCNT) , Kamand, India, 1–6. https://doi.org/10.1109/icccnt61001.2024.10725156 Kumlungmak K. Multi-agent deep reinforcement learning for cryptocurrency trading. Chula DigiVerse. 2024. https://doi.org/10.58837/chula.the.2022.95 . Lahmiri S, Bekiros S. Cryptocurrency forecasting with deep learning chaotic neural networks. Chaos Solitons Fractals. 2019;118:35–40. https://doi.org/10.1016/j.chaos.2018.11.014 . Lam KY, Makarov I. (2023). Financial Trading via Reinforcement Learning and Convolutional Neural Networks. 2023 IEEE 21st Jubilee International Symposium on Intelligent Systems and Informatics (SISY). https://doi.org/10.1109/sisy60376.2023.10417967 Leung MF, Chan L, Hung W-C, Tsoi S-F, Lam C, Cheng Y-H. (2023). An Intelligent System for Trading Signal of Cryptocurrency Based on Market Tweets Sentiments. FinTech . 2(1), 153–169. https://doi.org/10.3390/fintech2010011 Li X, Liu Y, Liu Z. Cryptocurrency return prediction: A machine learning analysis. Social Sci Res Netw. 2024. https://doi.org/10.2139/ssrn.4703167 . Lu J, Setthawong R, Setthawong P. (2024). A Framework Utilizing Genetic Algorithm Wrapped Neural Network Model in Discovering Effective Technical Indicators for the Cryptocurrency Market. 2024 8th International Conference on Information Technology (InCIT). https://doi.org/10.1109/incit63192.2024.10810651 Lua ZZ, Seow CK, Chan RCB, Cai Y, Cao Q. Automated Bitcoin Trading dApp Using Price Prediction from a Deep Learning Model. Risks. 2025;13(1):17. https://doi.org/10.3390/risks13010017 . Lucarelli G, Borrotti M. A Deep Reinforcement Learning Approach for Automated Cryptocurrency Trading. In: MacIntyre J, Maglogiannis I, Iliadis L, Pimenidis E, editors. Artificial Intelligence Applications and Innovations. AIAI 2019. IFIP Advances in Information and Communication Technology . Volume 559. Cham: Springer; 2019. https://doi.org/10.1007/978-3-030-19823-7_20 . Lucarelli G, Borrotti M. A deep Q-learning portfolio management framework for the cryptocurrency market. Neural Comput Applic. 2020;32:17229–44. https://doi.org/10.1007/s00521-020-05359-8 . Mai F, Shan Z, Bai Q, Wang X, Robert, Chiang RHL. How does social media impact Bitcoin value? A test of the silent majority hypothesis. J Manage Inform Syst. 2018;35(1):19–52. https://doi.org/10.1080/07421222.2018.1440774 . Majidi N, Shamsi M, Marvasti F. Algorithmic Trading Using Continuous Action Space Deep Reinforcement Learning. arXiv: Mach Learn. 2022. https://doi.org/10.48550/arXiv.2210.03469 . Mandych O, Staverska T, Maliy O. Integration of artificial intelligence into the blockchain and cryptocurrency market. Model Dev Economic Syst. 2023;2023(4). https://doi.org/10.31891/mdes/2023-10-8 . McNally S, Roche J, Caton S. (2018). Predicting the price of Bitcoin using machine learning. In 2018 26th Euromicro International Conference on Parallel, Distributed and Network-based Processing (PDP) (pp. 339–343). IEEE. https://doi.org/10.1109/PDP2018.2018.00060 Mikhaiylova SS, Sabirova SA. Development of a cryptocurrency trading strategy using machine learning methods. Comput Nanatechnol. 2024;11(2):11–21. https://doi.org/10.33693/2313-223x-2024-11-2-11-21 . Nakamoto S. (2008). Bitcoin: A peer-to-peer electronic cash system. Retrieved from https://bitcoin.org/bitcoin.pdf Okasová K, Košťál K. (2024). Using Machine Learning for Predicting Arbitrage Occurrences in Cryptocurrency Exchanges. 2024 IEEE International Conference on Blockchain and Cryptocurrency (ICBC) , Dublin, Ireland, 2024, 1–7. https://doi.org/10.1109/icbc59979.2024.10634339 Otabek S, Choi J. Multi-level deep Q-networks for Bitcoin trading strategies. Sci Rep. 2024;14:771. https://doi.org/10.1038/s41598-024-51408-w . Page MJ, McKenzie JE, Bossuyt PM, Boutron I, Hoffmann TC, Mulrow CD, Moher D. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ. 2021;372:n71. https://doi.org/10.1136/bmj.n71 . Pang Y, Sundararaj G, Ren J. (2019). Cryptocurrency Price Prediction using Time Series and Social Sentiment Data. BDCAT'19: Proceedings of the 6th IEEE/ACM International Conference on Big Data Computing, Applications and Technologies . 35–41. https://doi.org/10.1145/3365109.3368785 Park J, Seo Y-S. A Deep Learning-Based Action Recommendation Model for Cryptocurrency Profit Maximization. Electronics. 2022;11(9):1466. https://doi.org/10.3390/electronics11091466 . Pasupuleti MK. AI and the Future of Finance: Transforming Cryptocurrency, FinTech, and Economic Outlooks. Int J Acad Industrial Res Innovations(IJAIRI). 2024;4(4). https://doi.org/10.62311/nesx/97883 . Peng A, Ang SL, Lim CY. Automated Cryptocurrency Trading Bot Implementing DRL. Pertanika J Sci Technol. 2022;30(4). https://doi.org/10.47836/pjst.30.4.22 . Qing Y, Kong Y, Sun J, Lin J. (2022a). Cryptocurrency Trading Strategies and Profitability Based on Deep Learning Algorithms. 4th International Conference on Applied Machine Learning (ICAML). https://doi.org/10.1109/ICAML57167.2022.00079 Qing Y, Sun J, Kong Y, Lin J. (2022b). Fundamental Multi-factor Deep-learning Strategy For Cryptocurrency Trading. IEEE 20th International Conference on Industrial Informatics (INDIN) . https://doi.org/10.1109/INDIN51773.2022.9976116 Ramkumar G. (2021). Cryptocurrency Portfolio Construction Using Machine Learning Models. In: Jajuga, K., Locarek-Junge, H., Orlowski, L.T., Staehr, K, editors Contemporary Trends and Challenges in Finance. Springer Proceedings in Business and Economics . Springer, Cham. https://doi.org/10.1007/978-3-030-73667-5_7 Sanju T, Liyanage H, Bandara K, Kandakkulama D, Silva DD, Perera J. Stock-Crypto-App – Recommendation System for Stock and Cryptocurrency Market Using Cutting Edge Machine Learning Technology. Int Res J Innovations Eng Technol. 2023;7(9):110–7. https://doi.org/10.47001/irjiet/2023.709012 . Sattarov O, Muminov A, Lee CW, Kang HK, Oh R, Ahn J, Oh HJ, Jeon HS. Recommending cryptocurrency trading points with deep reinforcement learning approach. Appl Sci. 2020;10(4):1506. https://doi.org/10.3390/APP10041506 . Sebastião H, Godinho P. Forecasting and trading cryptocurrencies with machine learning under changing market conditions. Financial Innov. 2021;7(3). https://doi.org/10.1186/S40854-020-00217-X . Singh J, Thulasiram RK, Thavaneswaran A. (2022). LSTM based Algorithmic Trading model for Bitcoin. IEEE Symposium Series on Computational Intelligence . 344–351 https://doi.org/10.1109/SSCI51031.2022.10022021 Ślepaczuk R, Zenkova M. Robustness of Support Vector Machines in Algorithmic Trading on Cryptocurrency Market. Cent Eur Economic J. 2019;5(52). https://doi.org/10.1515/CEEJ-2018-0022 . Song D. Machine Learning for Price Prediction and Risk-Adjusted Portfolio Optimization in Cryptocurrencies. Mach Learn Model Techniques Financial Data Sci. 2025. https://doi.org/10.4018/979-8-3693-8186-1.ch013 . Srivastava V, Sikroria R. Ai and algorithmic trading: a study on predictive accuracy and market efficiency in fintech applications. ShodhKosh J Visual Perform Arts. 2024;5(1). https://doi.org/10.29121/shodhkosh.v5.i1.2024.2797 . Sruthi PJ, Shahithabanu M. AI Applications in Analysing and Predicting Cryptocurrency Market. Recent Trends Manage Commer. 2024;5(2):42–6. https://doi.org/10.46632/rmc/5/2/8 . Su PP. Adaptive Hybrid Optimized LSTM Models: A Novel Computational Framework for Algorithmic Financial Trading System. J Industrial Eng Appl Sci (JIEAS). 2024;2(6). https://doi.org/10.70393/6a69656173.323338 . Tanwar S, Raboaca MS. Stochastic Neural Networks-Based Algorithmic Trading for the Cryptocurrency Market. Mathematics. 2022;10(9):1456. https://doi.org/10.3390/math10091456 . Tapscott D, Tapscott A. Blockchain revolution: How the technology behind bitcoin is changing money, business, and the world. Penguin; 2016. Toleti J, Singh S, Kumar RP. (2024). Developing Robust Crypto-Currency Trading Strategies: A Fusion of Machine Learning and Hidden Markov Modelling for Enhanced Performance. IEEE International Conference on Electronics, Computing and Communication Technologies (CONECCT) , Bangalore, India, 1–6. https://doi.org/10.1109/conecct62155.2024.10677046 Viéitez A, Santos M, Naranjo R. Machine learning Ethereum cryptocurrency prediction and knowledge-based investment strategies. Knowl Based Syst. 2024;299:112088. https://doi.org/10.1016/j.knosys.2024.112088 . Yang Y, Zhang Z, Wang L. Cryptocurrency price prediction using deep learning and attention mechanism. PeerJ Comput Sci. 2020;6:e279. https://doi.org/10.7717/peerj-cs.279 . Żbikowski K. Application of Machine Learning Algorithms for Bitcoin Automated Trading. In: Ryżko D, Gawrysiak P, Kryszkiewicz M, Rybiński H, editors. Machine Intelligence and Big Data in Industry. Studies in Big Data. 19 ed. Cham: Springer; 2016. https://doi.org/10.1007/978-3-319-30315-4_14 . Zhao F, Zhang M, Zhou S, Lou Q. Application of Deep Reinforcement Learning for Cryptocurrency Market Trend Forecasting and Risk Management. J Industrial Eng Appl Sci (JIEAS). 2024;2(5). https://doi.org/10.5281/zenodo.13854906 . Zwitter A, Boisse-Despiaux M, Posada A. Blockchain for humanitarian action and development aid. J Int Humanitarian Action. 2020;5:16. https://doi.org/10.1186/s41018-020-00072-3 . Additional Declarations No competing interests reported. Supplementary Files Appendix.docx Cite Share Download PDF Status: Published Journal Publication published 13 Feb, 2026 Read the published version in Discover Artificial Intelligence → Version 1 posted Editorial decision: Revision requested 05 Nov, 2025 Reviews received at journal 22 Oct, 2025 Reviews received at journal 21 Oct, 2025 Reviews received at journal 20 Oct, 2025 Reviewers agreed at journal 15 Oct, 2025 Reviewers agreed at journal 14 Oct, 2025 Reviewers agreed at journal 13 Oct, 2025 Reviewers agreed at journal 13 Oct, 2025 Reviewers invited by journal 13 Oct, 2025 Editor invited by journal 30 Sep, 2025 Editor assigned by journal 28 Aug, 2025 Submission checks completed at journal 28 Aug, 2025 First submitted to journal 19 Aug, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7411123","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Systematic Review","associatedPublications":[],"authors":[{"id":533839207,"identity":"581fed6f-0ae1-4a2f-a46d-0e1bdf37bfe1","order_by":0,"name":"Péter Lengyel","email":"data:image/png;base64,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","orcid":"","institution":"University of Debrecen","correspondingAuthor":true,"prefix":"","firstName":"Péter","middleName":"","lastName":"Lengyel","suffix":""},{"id":533839208,"identity":"97758a6b-a673-4ae4-8e22-4098578a0d03","order_by":1,"name":"János Pancsira","email":"","orcid":"","institution":"University of Debrecen","correspondingAuthor":false,"prefix":"","firstName":"János","middleName":"","lastName":"Pancsira","suffix":""},{"id":533839209,"identity":"ade48983-862b-44b3-9ed0-8fad9a5d332a","order_by":2,"name":"István Füzesi","email":"","orcid":"","institution":"University of Debrecen","correspondingAuthor":false,"prefix":"","firstName":"István","middleName":"","lastName":"Füzesi","suffix":""}],"badges":[],"createdAt":"2025-08-19 18:23:18","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7411123/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7411123/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1007/s44163-025-00785-w","type":"published","date":"2026-02-13T15:58:09+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":94451471,"identity":"5545563d-fc0a-4982-a007-04ebbf060aa9","added_by":"auto","created_at":"2025-10-27 14:40:09","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":160224,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.docx","url":"https://assets-eu.researchsquare.com/files/rs-7411123/v1/3b95489afe5a35d760506f70.docx"},{"id":94452026,"identity":"f08f59fa-b4e9-4676-bec6-75786e55f80c","added_by":"auto","created_at":"2025-10-27 14:40:38","extension":"json","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":5386,"visible":true,"origin":"","legend":"","description":"","filename":"96d1e9f74dc44a55acf843d02a37de6b.json","url":"https://assets-eu.researchsquare.com/files/rs-7411123/v1/65d916fc15134311e41694dc.json"},{"id":94451852,"identity":"1c11128f-a76b-4fe2-be58-02330934f668","added_by":"auto","created_at":"2025-10-27 14:40:27","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":27885,"visible":true,"origin":"","legend":"","description":"","filename":"Appendix.docx","url":"https://assets-eu.researchsquare.com/files/rs-7411123/v1/61c76649b08bf644e9cef6cb.docx"},{"id":94451407,"identity":"ae5225f5-5469-43ae-b46f-a7865474f672","added_by":"auto","created_at":"2025-10-27 14:40:06","extension":"docx","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":15671,"visible":true,"origin":"","legend":"","description":"","filename":"Titlepage.docx","url":"https://assets-eu.researchsquare.com/files/rs-7411123/v1/12d4e5ba60dc4448203b4fb8.docx"},{"id":94451700,"identity":"22b60304-abd5-40c1-8e58-620da0866115","added_by":"auto","created_at":"2025-10-27 14:40:17","extension":"xml","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":169228,"visible":true,"origin":"","legend":"","description":"","filename":"96d1e9f74dc44a55acf843d02a37de6b1enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-7411123/v1/bcefe710af4a563e25810270.xml"},{"id":94450636,"identity":"cfb754a3-fc31-4720-92ad-5b5b7742934b","added_by":"auto","created_at":"2025-10-27 14:39:26","extension":"png","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":16437,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7411123/v1/813f7fedc772d275b0dfd005.png"},{"id":94451065,"identity":"19341657-9f75-41f4-8f95-71b5d9c2e9e5","added_by":"auto","created_at":"2025-10-27 14:39:46","extension":"png","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":20508,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7411123/v1/f0105bc957b7666d48e24622.png"},{"id":94451642,"identity":"00181766-08c4-44fb-bcb6-eb5e3e544936","added_by":"auto","created_at":"2025-10-27 14:40:15","extension":"xml","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":167525,"visible":true,"origin":"","legend":"","description":"","filename":"96d1e9f74dc44a55acf843d02a37de6b1structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-7411123/v1/ca27f327132bb0c2ddc97d8b.xml"},{"id":94451145,"identity":"1db6f07b-3475-42d5-aa4f-c7baa6a2a049","added_by":"auto","created_at":"2025-10-27 14:39:52","extension":"html","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":178142,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-7411123/v1/b08bef842cc3a881ce7b50c0.html"},{"id":94451710,"identity":"37a5ff49-b44e-43ea-8936-ce0da33024d2","added_by":"auto","created_at":"2025-10-27 14:40:18","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":38418,"visible":true,"origin":"","legend":"\u003cp\u003ePRISMA 2020 Flow Diagram for the Study Selection Process.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eSource: Own editing based on Page et al. (2021)\u003c/em\u003e\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7411123/v1/698c8d22d7a9d61a21ed5a0b.png"},{"id":94451813,"identity":"ec4f106c-c895-434d-9504-07304f84f54d","added_by":"auto","created_at":"2025-10-27 14:40:25","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":60263,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eVisual comparison of machine learning models used in cryptocurrency trading based on six key evaluation criteria (1–5 scale).\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eSource: Authors’ compilation based on reviewed articles\u003c/em\u003e\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7411123/v1/8d93575d2c876b2c02e2224b.png"},{"id":102786333,"identity":"4d142f47-ce56-40ce-8470-88ad2a0dcc69","added_by":"auto","created_at":"2026-02-16 16:12:47","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1349727,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7411123/v1/ad4c8268-dea0-4173-a57b-0589ead63b5d.pdf"},{"id":94451326,"identity":"00d78a45-0d9c-474e-91ae-51666d9214af","added_by":"auto","created_at":"2025-10-27 14:40:02","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":27885,"visible":true,"origin":"","legend":"","description":"","filename":"Appendix.docx","url":"https://assets-eu.researchsquare.com/files/rs-7411123/v1/0862e98c8c105ebf57fa41ef.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Machine Learning Integration in Cryptocurrency Trading: A Systematic Review of Fintech Implications","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eSince the emergence of Bitcoin in 2008, cryptocurrency markets have evolved into a global financial ecosystem, characterised by extreme volatility, decentralisation, and technological innovation (Nakamoto, \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2008\u003c/span\u003e; Feng et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). These markets challenge traditional financial models due to their non-linear dynamics and speculative behaviour, making them an ideal environment for the application of machine learning (ML) techniques (Kristjanpoller \u0026amp; Minutolo, \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2018\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eML has increasingly been adopted in cryptocurrency trading for its ability to process vast amounts of data, identify patterns, and adapt to evolving market conditions. Techniques such as supervised learning, deep learning, and reinforcement learning (RL) offer promising alternatives to rule-based strategies (McNally et al., \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Alessandretti et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). Among these, Long Short-Term Memory (LSTM) networks and Gated Recurrent Units (GRU) have proven effective in time series forecasting, while ensemble models such as Random Forest and XGBoost demonstrate robustness against noise (Sebasti\u0026atilde;o \u0026amp; Godinho, \u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Lahmiri \u0026amp; Bekiros, \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2019\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eReinforcement learning approaches like Deep Q-Networks (DQN) and Proximal Policy Optimization (PPO) are particularly suited for trading applications due to their ability to model sequential decision-making and maximise cumulative returns (Yang et al., \u003cspan citationid=\"CR67\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). Furthermore, hybrid models combining sentiment analysis with price prediction have gained attention, leveraging data from social media platforms to enhance prediction accuracy (Mai et al., \u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e2018\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eDespite these advances, several limitations remain. Overfitting, lack of interpretability, and insufficient generalisability are persistent technical concerns (Ji et al., \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). Ethical and regulatory challenges, such as algorithmic bias, data privacy, and market manipulation, also pose significant barriers to widespread adoption (Tapscott \u0026amp; Tapscott, \u003cspan citationid=\"CR64\" class=\"CitationRef\"\u003e2016\u003c/span\u003e; Zwitter et al., \u003cspan citationid=\"CR70\" class=\"CitationRef\"\u003e2020\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eThis study aims to systematically evaluate the integration of ML into cryptocurrency trading from a fintech perspective. By analysing 57 peer-reviewed studies, we assess the effectiveness of various ML models, examine risk mitigation practices, and identify both opportunities and challenges in practical implementation. The goal is to provide a structured understanding of how ML contributes to performance optimisation, ethical automation, and technological transformation in crypto-financial systems.\u003c/p\u003e"},{"header":"2. Methodology","content":"\u003cp\u003eThis study follows a systematic literature review approach to examine the application of machine learning (ML) in cryptocurrency trading and its implications for fintech. The methodology was designed to ensure comprehensive coverage, transparency, and replicability, following recognised best practices in systematic research.\u003c/p\u003e\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003ch2\u003e2.1 Research Questions\u003c/h2\u003e\u003cp\u003eThe review is structured around the following four research questions (RQs):\u003c/p\u003e\u003cp\u003e\u003cul\u003e\u003cli\u003e\u003cp\u003e\u003cb\u003eRQ1\u003c/b\u003e: How do ML algorithms enhance predictive accuracy and trading performance in cryptocurrency markets?\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003e\u003cb\u003eRQ2\u003c/b\u003e: What risk management techniques are facilitated by ML-driven trading strategies?\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003e\u003cb\u003eRQ3\u003c/b\u003e: How are these algorithms integrated into fintech infrastructures and automated trading systems?\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003e\u003cb\u003eRQ4\u003c/b\u003e: What ethical, regulatory, and practical limitations emerge in the use of ML for cryptocurrency trading?\u003c/p\u003e\u003c/li\u003e\u003c/ul\u003e\u003c/p\u003e\u003cp\u003eThese questions guided both the search strategy and the data synthesis stages.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec4\" class=\"Section2\"\u003e\u003ch2\u003e2.2 Literature Search and Study Selection\u003c/h2\u003e\u003cp\u003eA comprehensive search strategy was applied across major academic databases, including Scopus, Web of Science, IEEE Xplore, SpringerLink, and arXiv. The initial query, \u003cem\u003efintech implications of integrating ML in cryptocurrency trading\u003c/em\u003e, was expanded into focused expressions such as:\u003c/p\u003e\u003cp\u003e\u003cul\u003e\u003cli\u003e\u003cp\u003eReinforcement learning for portfolio optimisation in crypto markets\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003eML-based risk mitigation in crypto algorithmic trading\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003eSentiment analysis and price prediction in blockchain assets\u003c/p\u003e\u003c/li\u003e\u003c/ul\u003e\u003c/p\u003e\u003cp\u003eThe search covered peer-reviewed articles and high-quality preprints published between 2016 and 2025, limited to English-language studies directly addressing cryptocurrency trading.\u003c/p\u003e\u003cp\u003e\u003cb\u003eInclusion criteria\u003c/b\u003e:\u003c/p\u003e\u003cp\u003e\u003cul\u003e\u003cli\u003e\u003cp\u003eStudies applying supervised, unsupervised, deep, or reinforcement learning to cryptocurrency markets;\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003ePapers reporting empirical results on prediction, trading performance, or risk control;\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003eArticles discussing fintech-relevant elements such as automation, blockchain, or regulation.\u003c/p\u003e\u003c/li\u003e\u003c/ul\u003e\u003c/p\u003e\u003cp\u003e\u003cb\u003eExclusion criteria\u003c/b\u003e:\u003c/p\u003e\u003cp\u003e\u003cul\u003e\u003cli\u003e\u003cp\u003eStudies on traditional finance only (e.g., stock or forex markets);\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003eNon-empirical works or those without reproducible methods;\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003eDuplicate or off-topic publications.\u003c/p\u003e\u003c/li\u003e\u003c/ul\u003e\u003c/p\u003e\u003cp\u003eThe screening and selection process adhered to the PRISMA 2020 guidelines (Page et al., \u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e2021\u003c/span\u003e), ensuring transparent and systematic identification, screening, and reporting. Of the initial set of 282 articles, 57 were retained after title/abstract screening and full-text evaluation. Citation chaining (both backward and forward) was used to supplement the final sample (see Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eThe final set of articles was selected through a transparent screening pipeline using PRISMA principles, complemented by best practices in systematic data collection and visualisation (Bohr \u0026amp; Memarzadeh, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2020\u003c/span\u003e).\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003cem\u003eSource: Own editing based on\u003c/em\u003e Page et al. (\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e2021\u003c/span\u003e)\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec5\" class=\"Section2\"\u003e\u003ch2\u003e2.3 Data Extraction and Thematic Analysis\u003c/h2\u003e\u003cp\u003eData were extracted along five key dimensions:\u003c/p\u003e\u003cp\u003e\u003col\u003e\u003cspan\u003e\u003cli\u003e\u003cp\u003e\u003cb\u003eModel characteristics\u003c/b\u003e: algorithm type (e.g., LSTM, GRU, DQN), learning framework, optimisation goal;\u003c/p\u003e\u003c/li\u003e\u003c/span\u003e\u003cspan\u003e\u003cli\u003e\u003cp\u003e\u003cb\u003ePerformance metrics\u003c/b\u003e: accuracy, F1 score, RMSE, Sharpe ratio, ROI;\u003c/p\u003e\u003c/li\u003e\u003c/span\u003e\u003cspan\u003e\u003cli\u003e\u003cp\u003e\u003cb\u003eRisk management methods\u003c/b\u003e: Value at Risk (VaR), stop-loss strategies, reward shaping;\u003c/p\u003e\u003c/li\u003e\u003c/span\u003e\u003cspan\u003e\u003cli\u003e\u003cp\u003e\u003cb\u003eComputational aspects\u003c/b\u003e: latency, backtesting, scalability;\u003c/p\u003e\u003c/li\u003e\u003c/span\u003e\u003cspan\u003e\u003cli\u003e\u003cp\u003e\u003cb\u003eFintech integration\u003c/b\u003e: blockchain infrastructure, DeFi mechanisms, smart contracts, and regulatory concerns.\u003c/p\u003e\u003c/li\u003e\u003c/span\u003e\u003c/ol\u003e\u003c/p\u003e\u003cp\u003eThematic analysis was conducted using Braun and Clarke\u0026rsquo;s (\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2006\u003c/span\u003e) six-phase approach. This combined quantitative aggregation (e.g., number of studies using RL models) with qualitative coding (e.g., ethical implications, automation benefits).\u003c/p\u003e\u003cp\u003eEach study was mapped to one or more of the research questions. Summarised findings are presented in the Results section, while detailed tabular comparisons are provided in Appendix A to maintain clarity and conciseness in the main body of the paper.\u003c/p\u003e\u003c/div\u003e"},{"header":"3. Results","content":"\u003cdiv id=\"Sec7\" class=\"Section2\"\u003e\u003ch2\u003e3.1. Descriptive Overview of Reviewed ML-Based Crypto Trading Studies\u003c/h2\u003e\u003cp\u003eThis section maps the research landscape of the literature on fintech implications of integrating machine learning algorithms into cryptocurrency trading strategies, encompassing a broad spectrum of methodologies, including supervised learning, reinforcement learning, deep learning, and hybrid approaches. The studies predominantly focus on predictive modeling for price forecasting, trading strategy optimization, and risk management within volatile cryptocurrency markets, often leveraging neural networks, decision trees, and reinforcement learning frameworks. Geographic and disciplinary diversity is evident, with contributions from computational finance, artificial intelligence, and ethical regulatory perspectives, reflecting the multifaceted nature of fintech innovation. This comparative analysis addresses key research questions by synthesizing empirical findings on algorithmic performance, risk mitigation, and the broader fintech ecosystem impacts, including ethical and regulatory considerations.\u003c/p\u003e\u003cp\u003eTo provide a comprehensive view of the literature, the following part of this subsection presents a thematic analysis structured around five key domains identified during the review: predictive accuracy, trading performance, risk management effectiveness, algorithmic efficiency, and fintech integration impact. This classification reflects the dominant methodological and conceptual trends observed across the selected articles.\u003c/p\u003e\u003cp\u003eA detailed summary of all 57 reviewed studies, including key characteristics, applied machine learning techniques, and major findings, is presented in Appendix A (Appendix A. Summary Table of Reviewed Studies on ML in Cryptocurrency Trading). The appendix table is organised along the same five analytical dimensions, thus reinforcing the thematic synthesis and enhancing the transparency and reproducibility of this systematic review.\u003c/p\u003e\u003cp\u003e\u003cb\u003ePredictive Accuracy\u003c/b\u003e:\u003c/p\u003e\u003cp\u003eNumerous studies (n\u0026thinsp;=\u0026thinsp;40) reported that deep learning architectures such as LSTM and GRU offer substantial improvements in prediction accuracy over traditional models (Lu et al., \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Vi\u0026eacute;itez et al., \u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Toleti et al., \u003cspan citationid=\"CR65\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Ensemble methods and reinforcement learning strategies further enhance robustness across different market regimes (Huang, \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Jabbar \u0026amp; Jalil, \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Bouteska et al., \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Moreover, integrating sentiment analysis from social media and causal feature engineering boosts performance (Belcastro et al., \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Avramelou et al., \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Amirzadeh et al., \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2023\u003c/span\u003e).\u003c/p\u003e\u003cp\u003e\u003cb\u003eTrading Performance\u003c/b\u003e:\u003c/p\u003e\u003cp\u003eA total of 38 studies indicated that ML-based trading strategies outperformed traditional benchmarks like buy-and-hold, especially in terms of Sharpe ratio, return consistency, and drawdown reduction (Lua et al., \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Otabek \u0026amp; Choi, \u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Asgari \u0026amp; Khasteh, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). Reinforcement learning models with novel reward structures yielded notable adaptability and profitability (Kang et al., \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Kouloumpris \u0026amp; Vlahavas, \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Lucarelli \u0026amp; Borrotti, \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). Some researchers focused on portfolio optimisation and risk-adjusted return frameworks (Song, \u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Ramkumar, \u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Lucarelli et al., 2020).\u003c/p\u003e\u003cp\u003e\u003cb\u003eRisk Management Effectiveness\u003c/b\u003e:\u003c/p\u003e\u003cp\u003eThirty studies incorporated financial risk metrics\u0026mdash;such as Value at Risk (VaR), drawdown controls, and the Sharpe Ratio, to evaluate and mitigate model-induced risks (Huang, \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Zhao et al., \u003cspan citationid=\"CR69\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Ślepaczuk \u0026amp; Zenkova, \u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). Hybrid architectures using technical indicators in conjunction with ML methods enhanced the detection of market anomalies and reduced false trading signals (Tanwar \u0026amp; Raboaca, \u003cspan citationid=\"CR63\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Bellocca et al., \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Reinforcement learning models featuring penalty mechanisms and conservative position sizing demonstrated prudent risk-reward trade-offs (Amirzadeh et al., \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Kumlungmak, \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2024\u003c/span\u003e).\u003c/p\u003e\u003cp\u003e\u003cb\u003eAlgorithmic Efficiency\u003c/b\u003e:\u003c/p\u003e\u003cp\u003eApproximately 25 studies dealt with the practical implementation of ML models, focusing on scalability, real-time performance, and computational efficiency. Genetic algorithms and hyperparameter optimisation techniques were deployed to reduce overhead and improve performance (Lu et al., \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Su, \u003cspan citationid=\"CR62\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Okasov\u0026aacute; \u0026amp; Košť\u0026aacute;l, 2024). Deep reinforcement learning with attention mechanisms proved effective in multi-asset scenarios and long sequence processing (Betancourt \u0026amp; Chen, \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2021\u003c/span\u003e).\u003c/p\u003e\u003cp\u003e\u003cb\u003eFintech Integration Impact\u003c/b\u003e:\u003c/p\u003e\u003cp\u003eAround 20 studies tackled the broader fintech implications of ML integration. Key themes included regulatory compliance, explainability, fairness, and AI governance in crypto-finance ecosystems (Kumari et al., \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Pasupuleti, \u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Alibašić, \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). The convergence of AI and blockchain improved system transparency, fraud detection, and investor trust (Mandych et al., \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Hou, \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). Despite these gains, the literature identified a need for consistent ethical and regulatory frameworks to ensure responsible deployment (Srivastava \u0026amp; Sikroria, \u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e2024\u003c/span\u003e).\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\u003ch2\u003e3.2. Temporal Trends in ML Applications for Cryptocurrency Trading\u003c/h2\u003e\u003cp\u003eThe integration of machine learning into cryptocurrency trading strategies has evolved considerably from 2016 onwards, reflecting increasing sophistication in modeling techniques and expanding fintech implications. Early research primarily focused on applying basic machine learning models to price prediction and algorithmic trading in cryptocurrencies. Over time, studies incorporated deep learning and reinforcement learning to tackle market volatility and enhance decision-making in automated trading systems. Recent advancements emphasize hybrid models, sentiment analysis, ethical considerations, and the development of robust frameworks to optimize trading performance while addressing regulatory and operational challenges in the dynamic cryptocurrency market. A structured summary is provided in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e2\u003c/span\u003e.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eChronological Evolution of ML Applications in Cryptocurrency Trading (2016\u0026ndash;2025)\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"3\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eYears\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eResearch Direction\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eDescription\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e2016\u0026ndash;2018\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eEarly Machine Learning Applications in Cryptocurrency Trading\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eInitial studies applied fundamental machine learning techniques such as support vector machines and neural networks for price prediction and trading signal generation. Research focused on exploiting market inefficiencies and developing algorithmic trading strategies that outperform traditional benchmarks, with preliminary insights into market sentiment and social media data integration.\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e2019\u0026ndash;2020\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eEmergence of Deep Reinforcement Learning and Portfolio Optimization\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eResearch introduced deep reinforcement learning methods and advanced portfolio management frameworks to enhance profitability and risk-adjusted returns in cryptocurrency trading. Studies explored Q-learning variants and actor-critic models, integrating technical indicators and sentiment analysis to improve trading decisions and strategy robustness.\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e2021\u0026ndash;2022\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eAdvancement of Hybrid and Deep Learning Models for Automated Trading\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eFocus shifted toward combining deep learning architectures like LSTM, CNN, and self-attention networks with reinforcement learning to develop adaptive and high-frequency trading algorithms. Emphasis was placed on improving predictive accuracy, managing market risks, and leveraging multi-factor models alongside social sentiment for enhanced trading performance.\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e2023\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eIntegration of Causal Analysis, Ethical Frameworks, and Sentiment-Aware Models\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eResearch expanded to include causal analysis frameworks to empower reinforcement learning agents and the ethical implications of AI and ML use in cryptocurrency trading. Sentiment-aware deep reinforcement learning models and safety mechanisms were proposed to mitigate risk and improve decision-making under volatile market conditions. Ethical considerations and regulatory compliance became central themes alongside technical innovations.\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e2024\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eOptimization of Machine Learning Frameworks and Fintech Innovations\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eStudies emphasized the optimization of genetic algorithms, hybrid models, and multi-agent reinforcement learning approaches to address market volatility and arbitrage opportunities. The integration of smart contracts and blockchain oracles with ML models facilitated transparent and automated portfolio management. Research also benchmarked different ML algorithms for their effectiveness in volatility prediction, risk adjustment, and profitability across diverse cryptocurrencies.\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e2025\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eDeployment of Advanced ML Systems in Real-World Trading Environments\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eThe latest research demonstrated that the deployment of ML-powered decentralized applications and trading bots achieves substantial returns on investment. Emphasis was placed on practical implementations of LSTM, gradient boosting, and reinforcement learning models within secure and trust-free fintech infrastructures, highlighting scalability, transparency, and adaptability in real-time cryptocurrency trading.\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003e\u003cstrong\u003eSource\u003c/strong\u003e\u003cp\u003eAuthor\u0026rsquo;s own editing based on reviewed articles\u003c/p\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e\u003ch2\u003e3.3. Thematic Insights into ML Strategies and Fintech Integration\u003c/h2\u003e\u003cp\u003eThe literature on integrating machine learning algorithms into cryptocurrency trading strategies reveals several dominant themes focused on enhancing trading performance, risk management, and fintech innovation. A large proportion of studies emphasize predictive modeling and algorithmic trading strategies employing diverse machine learning techniques such as deep learning, reinforcement learning, and ensemble methods. Significant attention is also given to managing the high volatility and risk inherent in cryptocurrency markets through advanced methods, including hybrid models and risk-adjusted portfolio optimization. Emerging research highlights ethical, regulatory, and operational challenges, underscoring the need for frameworks that ensure responsible AI use while fostering financial innovation in this rapidly evolving fintech landscape. The classification is summarised in Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e3\u003c/span\u003e.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eDominant Research Themes in ML-based Cryptocurrency Trading (n\u0026thinsp;=\u0026thinsp;57 papers)\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"3\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTheme\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eAppears In\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eTheme Description\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMachine Learning Models for Price Prediction and Trading Performance\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e43 Papers\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eExtensive research focuses on employing machine learning models such as neural networks (LSTM, GRU), random forests, support vector machines, and ensemble methods to predict cryptocurrency prices and inform trading strategies. These models have demonstrated improved predictive accuracy and profitability over traditional benchmarks across various cryptocurrencies under diverse market conditions (Lu et al., \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Huang, \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Vi\u0026eacute;itez et al., \u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Song, \u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Jabbar \u0026amp; Jalil, \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Hafid et al., \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Park \u0026amp; Seo, \u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Bouteska et al., \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Sebasti\u0026atilde;o \u0026amp; Godinho, \u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Singh et al., \u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Li et al., \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2024\u003c/span\u003e).\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eReinforcement Learning for Algorithmic Trading and Portfolio Optimization\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e24 Papers\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eReinforcement learning (RL) techniques, including Deep Q-networks, Advantage Actor-Critic, Proximal Policy Optimization, and multi-agent frameworks, are widely applied to develop adaptive trading strategies that optimize asset allocation and manage risk. Studies show RL's capability to handle dynamic market environments and improve returns while balancing risk, especially in volatile cryptocurrency markets (Kang et al., \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Otabek \u0026amp; Choi, \u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Kouloumpris \u0026amp; Vlahavas, \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Amirzadeh et al., \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Zhao et al., \u003cspan citationid=\"CR69\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Peng et al., \u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Asgari \u0026amp; Khasteh, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Felizardo et al., \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Betancourt \u0026amp; Chen, \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Lucarelli et al., 2020; Bu \u0026amp; Cho, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Lucarelli \u0026amp; Borrotti, \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Sattarov et al., \u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Kumlungmak, \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2024\u003c/span\u003e).\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRisk Management and Volatility Mitigation Using Machine Learning\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e23 Papers\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eAddressing cryptocurrency market volatility and associated risks is a key focus, with machine learning enabling advanced risk mitigation techniques such as EGARCH integration, stochastic neural networks, hybrid models, and portfolio optimization frameworks that incorporate risk-adjusted metrics like Sharpe ratio and Value at Risk. These approaches enhance trading robustness and stability (Lu et al., \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Song, \u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Kouloumpris \u0026amp; Vlahavas, \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Zhao et al., \u003cspan citationid=\"CR69\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Tanwar \u0026amp; Raboaca, \u003cspan citationid=\"CR63\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Alaminos et al., \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Chen, \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Koker \u0026amp; Koutmos, \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2020\u003c/span\u003e).\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eIntegration of Sentiment Analysis and External Data Sources\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e14 Papers\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eSeveral studies incorporate sentiment analysis from social media and market indicators to enhance price prediction models. The integration of natural language processing and sentiment scoring improves understanding of market dynamics and contributes to more informed trading decisions, although the impact varies across models and cryptocurrencies (Vi\u0026eacute;itez et al., \u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Belcastro et al., \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Avramelou et al., \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Leung et al., \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Pang et al., \u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e2019\u003c/span\u003e).\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eEthical, Regulatory, and Operational Implications of AI in Cryptocurrency Trading\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e9 Papers\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eEmerging literature explores ethical frameworks and regulatory challenges posed by AI and machine learning in cryptocurrency trading, emphasizing the need for responsible AI deployment, transparency, market integrity, investor protection, and novel regulatory approaches tailored to digital assets\u0026rsquo; unique characteristics (Kumari et al., \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Srivastava \u0026amp; Sikroria, \u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Pasupuleti, \u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Mandych et al., \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Alibašić, \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2023\u003c/span\u003e).\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eHybrid and Ensemble Methodologies for Enhanced Trading Strategies\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e12 Papers\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eHybrid models combining deep learning architectures (e.g., CNN-LSTM, LSTM-GRU) and ensemble learning methods have been proposed to leverage complementary strengths of different algorithms, improving predictive performance and trading profitability in volatile cryptocurrency markets (Park \u0026amp; Seo, \u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Bouteska et al., \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Bellocca et al., \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Qing et al., \u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e2022a\u003c/span\u003e).\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAutomated Trading Systems and Decentralized Finance Integration\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e8 Papers\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eResearch includes blockchain and smart contract integration with machine learning-based trading systems, enhancing transparency, automation, and security in cryptocurrency trading. These systems reduce barriers for investors and enable decentralized portfolio management (Lua et al., \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Mandych et al., \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Hou, \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2025\u003c/span\u003e).\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eArbitrage Detection and Statistical Trading Techniques Using Machine Learning\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e4 Papers\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eMachine learning models are utilized to identify arbitrage opportunities and apply statistical arbitrage trading strategies in cryptocurrency markets, offering potential for riskless profits but facing challenges due to rapid market changes (Okasov\u0026aacute; \u0026amp; Košť\u0026aacute;l, 2024; Fischer et al., \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2019\u003c/span\u003e).\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003e\u003cstrong\u003eSource\u003c/strong\u003e\u003cp\u003eAuthor\u0026rsquo;s own editing based on reviewed articles\u003c/p\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec10\" class=\"Section2\"\u003e\u003ch2\u003e3.4. Cross-Study Evaluation of Methodologies and Fintech Implications\u003c/h2\u003e\u003cp\u003eThe reviewed literature on the fintech implications of integrating ML algorithms into cryptocurrency trading strategies reveals a dynamic and rapidly evolving field marked by innovative methodologies and promising empirical results. A key theme is the demonstrated potential of ML models to enhance predictive accuracy, trading profitability, and risk management in highly volatile cryptocurrency markets. However, the diversity of approaches and varying data quality pose challenges to generalizability and robustness. Additionally, ethical, regulatory, and operational considerations remain underexplored, highlighting critical gaps for future research. This synthesis critically evaluates the strengths and limitations across methodological rigor, algorithmic performance, risk mitigation, fintech integration, and ethical/regulatory discourse.\u003c/p\u003e\u003cp\u003eThe analysis is based on the full set of 57 reviewed studies and is summarised in a comparative matrix provided in Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e4\u003c/span\u003e. The table includes study-specific evaluations aligned with each analytical category and highlights both strengths and limitations. This structured synthesis aims to guide future research by clarifying which areas exhibit robust evidence and where significant knowledge gaps remain.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 4\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eComparative Matrix of Strengths and Weaknesses in Reviewed Studies on Machine Learning for Cryptocurrency Trading\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"3\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAspect\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eStrengths\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eWeaknesses\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMethodological Approaches\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eThe literature employs a broad spectrum of ML techniques, including neural networks, reinforcement learning, genetic algorithms, and ensemble methods, demonstrating adaptability to complex market dynamics and non-linear patterns (Lu et al., \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Su, \u003cspan citationid=\"CR62\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Song, \u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Otabek \u0026amp; Choi, \u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Rigorous backtesting, cross-validation, and real-world trading simulations enhance the credibility of findings (Huang, \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Jabbar \u0026amp; Jalil, \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Sattarov et al., \u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). The integration of hybrid models, such as combining deep learning with Hidden Markov Models or causal analysis, further strengthens predictive capabilities (Toleti et al., \u003cspan citationid=\"CR65\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Amirzadeh et al., \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2023\u003c/span\u003e).\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eDespite methodological diversity, many studies rely heavily on historical data with limited temporal scope, which may not capture evolving market regimes or rare events (Vi\u0026eacute;itez et al., \u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Belcastro et al., \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). Some models exhibit overfitting risks due to high parameterization and insufficient out-of-sample validation (Srivastava \u0026amp; Sikroria, \u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Sebasti\u0026atilde;o \u0026amp; Godinho, \u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). The heterogeneity in data preprocessing and feature selection complicates direct comparison and replication (Sruthi \u0026amp; Shahithabanu, \u003cspan citationid=\"CR61\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Sebasti\u0026atilde;o \u0026amp; Godinho, \u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). Moreover, few studies address the impact of transaction costs and slippage comprehensively (Lam \u0026amp; Makarov, \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2023\u003c/span\u003e).\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAlgorithmic Performance and Predictive Accuracy\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eML models, particularly LSTM, GRU, Random Forest, and reinforcement learning algorithms like PPO and DDPG, consistently outperform traditional benchmarks such as buy-and-hold strategies, achieving significant returns and improved Sharpe ratios (Lu et al., \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Lua et al., \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Kang et al., \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Otabek \u0026amp; Choi, \u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Lucarelli \u0026amp; Borrotti, \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). The use of sentiment analysis and social media data enhances prediction robustness in some cases (Vi\u0026eacute;itez et al., \u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Belcastro et al., \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Avramelou et al., \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). Reinforcement learning frameworks demonstrate adaptability to market volatility and dynamic portfolio optimization (Kang et al., \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Zhao et al., \u003cspan citationid=\"CR69\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Kumlungmak, \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2024\u003c/span\u003e).\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003ePerformance gains are often context-specific, with some models showing diminished effectiveness during extreme market conditions or bearish phases (Sebasti\u0026atilde;o \u0026amp; Godinho, \u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Kumlungmak, \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Sentiment analysis integration yields mixed results, with limited incremental benefit in certain studies (Vi\u0026eacute;itez et al., \u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). The computational complexity and training time of deep models may hinder real-time applicability (Huang, \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Lam \u0026amp; Makarov, \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). Additionally, the lack of standardized evaluation metrics and inconsistent reporting of risk-adjusted returns limit comprehensive assessment (Jabbar \u0026amp; Jalil, \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Srivastava \u0026amp; Sikroria, \u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e2024\u003c/span\u003e).\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRisk Management and Volatility Handling\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eSeveral studies incorporate risk-adjusted metrics such as Sharpe ratio, Value at Risk (VaR), and drawdown controls, demonstrating ML\u0026rsquo;s capacity to balance profitability with risk mitigation (Huang, \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Kang et al., \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Zhao et al., \u003cspan citationid=\"CR69\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Koker \u0026amp; Koutmos, \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). Hybrid approaches combining technical indicators with ML models improve robustness against market noise and volatility (Lu et al., \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) (Kouloumpris \u0026amp; Vlahavas, \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2023\u003c/span\u003e) (Alaminos et al., \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Reinforcement learning agents are designed with reward functions and safety mechanisms to reduce exposure during uncertain periods (Kouloumpris \u0026amp; Vlahavas, \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2023\u003c/span\u003e).\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eRisk management strategies are often heuristic or model-specific, lacking universal frameworks applicable across different cryptocurrencies or market conditions (Kouloumpris \u0026amp; Vlahavas, \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Kumlungmak, \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Some reinforcement learning models prioritize profit maximization at the expense of risk awareness, leading to potential overexposure (Kouloumpris \u0026amp; Vlahavas, \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Asgari \u0026amp; Khasteh, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). The challenge of modeling extreme tail risks and sudden market shocks remains insufficiently addressed (Alaminos et al., \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Furthermore, the impact of leverage and liquidity constraints is rarely considered (Fischer et al., \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2019\u003c/span\u003e).\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eIntegration within Fintech Infrastructure\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eThe integration of ML models with blockchain technologies, smart contracts, and decentralized applications (dApps) enhances transparency, automation, and accessibility in cryptocurrency trading (Lua et al., \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2025\u003c/span\u003e) (Pasupuleti, \u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Mandych et al., \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). Studies highlight the potential for ML-driven portfolio management systems to reduce barriers for retail investors and improve operational efficiency (Lua et al., \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Ramkumar, \u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). The use of oracles and on-chain verification mechanisms supports real-time, trustless trading environments (Lua et al., \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2025\u003c/span\u003e).\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003ePractical deployment challenges include scalability, latency, and security vulnerabilities inherent in blockchain-based systems (Mandych et al., \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). The complexity of integrating ML predictions with decentralized finance (DeFi) protocols is underexplored (Pasupuleti, \u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Many studies remain theoretical or simulation-based without full implementation in live fintech ecosystems (Sruthi \u0026amp; Shahithabanu, \u003cspan citationid=\"CR61\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Hou, \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). Additionally, interoperability issues between ML models and existing financial infrastructures pose barriers (Pasupuleti, \u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e2024\u003c/span\u003e).\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eEthical, Regulatory, and Operational Considerations\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eEmerging research acknowledges the ethical implications of AI and ML in cryptocurrency trading, emphasizing the need for responsible frameworks to ensure market integrity and investor protection (Alibašić, \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). Discussions include the risks of algorithmic biases, market manipulation, and the necessity for adaptive regulatory approaches tailored to digital assets (Srivastava \u0026amp; Sikroria, \u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Mandych et al., \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Alibašić, \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). The potential for AI to enhance compliance through RegTech applications is recognized (Pasupuleti, \u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e2024\u003c/span\u003e).\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eEthical and regulatory aspects are insufficiently addressed in most empirical studies, with limited operational guidelines or frameworks proposed (Sruthi \u0026amp; Shahithabanu, \u003cspan citationid=\"CR61\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Alibašić, \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). The fast-paced evolution of ML-driven trading outstrips current regulatory capacities, raising concerns about transparency and accountability (Srivastava \u0026amp; Sikroria, \u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Mandych et al., \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). Operational challenges such as model interpretability, data privacy, and systemic risks remain largely unexplored (Alibašić, \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). The literature lacks consensus on best practices for balancing innovation with ethical constraints (Alibašić, \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2023\u003c/span\u003e).\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eData Quality and Feature Engineering\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eStudies leverage diverse data sources, including technical indicators, market prices, social media sentiment, and blockchain metrics, enriching model inputs and enhancing predictive power (Vi\u0026eacute;itez et al., \u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Belcastro et al., \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Avramelou et al., \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Li et al., \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Advanced feature selection and causal analysis improve model interpretability and relevance (Amirzadeh et al., \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Li et al., \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2024\u003c/span\u003e).\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eData limitations include short historical records, especially for newer cryptocurrencies, leading to potential biases and reduced model generalizability (Park \u0026amp; Seo, \u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Alaminos et al., \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). The noisy and unstructured nature of social media data challenges sentiment analysis accuracy (Vi\u0026eacute;itez et al., \u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Leung et al., \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). Inconsistent data preprocessing and lack of standardized datasets hinder reproducibility and benchmarking (Sruthi \u0026amp; Shahithabanu, \u003cspan citationid=\"CR61\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Avramelou et al., \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). The impact of data latency and real-time availability on model performance is rarely discussed (Lua et al., \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2025\u003c/span\u003e).\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eComparative Evaluation of ML Algorithms\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eComprehensive comparisons reveal that ensemble methods and deep learning architectures often outperform simpler models, with tree-based models like Random Forest excelling in return predictability (Jabbar \u0026amp; Jalil, \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) (Bouteska et al., \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) (Li et al., \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Reinforcement learning approaches show promise in dynamic strategy adaptation and portfolio management (Kang et al., \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Otabek \u0026amp; Choi, \u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Kumlungmak, \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Hybrid models combining multiple ML techniques achieve superior results (Toleti et al., \u003cspan citationid=\"CR65\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Amirzadeh et al., \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2023\u003c/span\u003e).\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eComparative studies are limited by inconsistent experimental setups, varying datasets, and differing evaluation metrics, complicating definitive conclusions (Jabbar \u0026amp; Jalil, \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) (Bouteska et al., \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Some algorithms require extensive hyperparameter tuning and computational resources, limiting practical deployment (Huang, \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Srivastava \u0026amp; Sikroria, \u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). The trade-offs between model complexity, interpretability, and real-time applicability are insufficiently explored (Felizardo et al., \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Lam \u0026amp; Makarov, \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). Few studies address the robustness of algorithms across diverse market regimes (Sebasti\u0026atilde;o \u0026amp; Godinho, \u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e2021\u003c/span\u003e).\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003e\u003cstrong\u003eSource\u003c/strong\u003e\u003cp\u003e\u003cem\u003eAuthor\u0026rsquo;s own editing based on reviewed articles\u003c/em\u003e\u003c/p\u003e\u003c/p\u003e\u003cp\u003eTo enhance the clarity of the comparative findings, a visual summary is provided in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e.\u003c/p\u003e\u003cp\u003eThe heatmap highlights the relative strengths and limitations of key machine learning models based on six evaluation criteria: predictive accuracy, return on investment, scalability, risk management capability, interpretability, and regulatory alignment.\u003c/p\u003e\u003cp\u003eThe numerical ratings (on a 1\u0026ndash;5 scale) are derived from the authors\u0026rsquo; synthesis of the reviewed literature and represent a qualitative aggregation of model characteristics discussed in Sections \u003cspan refid=\"Sec7\" class=\"InternalRef\"\u003e3.1\u003c/span\u003e to \u003cspan refid=\"Sec10\" class=\"InternalRef\"\u003e3.4\u003c/span\u003e.\u003c/p\u003e\u003cp\u003eThis visualisation complements Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e4\u003c/span\u003e and supports readers in quickly identifying which models are most suitable under specific constraints and objectives.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003cem\u003eSource: Authors\u0026rsquo; compilation based on reviewed articles\u003c/em\u003e\u003c/p\u003e\u003cp\u003eAs shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e, deep learning models such as LSTM and GRU demonstrate high predictive accuracy and solid ROI, but they generally lack interpretability and scalability. In contrast, traditional ensemble methods like Random Forest and XGBoost offer better scalability and interpretability, though at the cost of slightly lower predictive performance.\u003c/p\u003e\u003cp\u003eReinforcement learning (RL) techniques appear particularly effective in risk-sensitive environments, achieving high ROI and strong risk management capabilities, albeit with limitations in real-time scalability.\u003c/p\u003e\u003cp\u003eHybrid and ensemble models generally offer a balanced trade-off across most dimensions, but their complexity may hinder deployment in resource-constrained environments.\u003c/p\u003e\u003cp\u003eThese distinctions underscore the importance of aligning model choice with specific application goals, for example, using RL models in highly dynamic and volatile markets, whereas Random Forest may be more appropriate when interpretability and compliance requirements are prioritised.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\u003ch2\u003e3.5. Theoretical and Practical Implications\u003c/h2\u003e\u003cp\u003eTo bridge the gap between academic inquiry and real-world application, this section synthesises the theoretical contributions and practical implications of integrating machine learning (ML) into cryptocurrency trading. The reviewed literature indicates how advanced computational models not only reshape predictive finance theory but also provide actionable insights for traders, fintech developers, and policymakers.\u003c/p\u003e\u003cp\u003e\u003cb\u003eTheoretical Implications\u003c/b\u003e\u003c/p\u003e\u003cp\u003e\u003cul\u003e\u003cli\u003e\u003cp\u003eThe integration of machine learning algorithms into cryptocurrency trading strategies substantiates the hypothesis that ML can significantly enhance predictive accuracy and trading profitability. These findings support the view that advanced computational models outperform traditional financial models in volatile markets, thus challenging classical efficient market hypotheses by revealing exploitable inefficiencies in cryptocurrency markets (Alessandretti et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Jabbar \u0026amp; Jalil, \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Lu et al., \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2024\u003c/span\u003e).\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003eReinforcement learning (RL) and deep learning models, particularly those employing hybrid and multi-agent frameworks, contribute to theoretical developments by addressing dynamic market conditions and risk management. Their adaptability and robustness in non-stationary environments expand the theory of algorithmic trading under uncertainty (Asgari \u0026amp; Khasteh, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Kang et al., \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Kumlungmak, \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2024\u003c/span\u003e).\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003eThe incorporation of sentiment analysis and social media data into ML-based forecasting introduces a behavioural dimension to price prediction theories. This highlights the influence of investor sentiment and informational flows on cryptocurrency price dynamics, bridging finance, behavioural economics, and data science (Avramelou et al., \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Belcastro et al., \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Pang et al., \u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e2019\u003c/span\u003e).\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003eEthical and regulatory aspects arising from AI-driven trading challenge conventional frameworks of market efficiency and fairness. These findings suggest the need for expanded models that integrate the normative and institutional consequences of algorithmic decision-making in financial markets (Alibašić, \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Mandych et al., \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Srivastava \u0026amp; Sikroria, \u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e2024\u003c/span\u003e).\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003eComparative evaluations of ML algorithms indicate that ensemble and deep learning approaches, such as gradient boosting and LSTM networks, excel at capturing short- and long-term temporal dependencies in highly volatile price series. These insights refine theoretical models of time series forecasting and financial prediction (Alessandretti et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Bouteska et al., \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Li et al., \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2024\u003c/span\u003e).\u003c/p\u003e\u003c/li\u003e\u003c/ul\u003e\u003c/p\u003e\u003cp\u003e\u003cb\u003ePractical Implications\u003c/b\u003e\u003c/p\u003e\u003cp\u003e\u003cul\u003e\u003cli\u003e\u003cp\u003eThe consistent outperformance and robustness of ML-based trading strategies suggest considerable potential for fintech firms and individual traders to optimise portfolio management and risk mitigation. The findings support the broader application of AI-powered tools for real-time decision-making in digital asset markets (Jabbar \u0026amp; Jalil, \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Lua et al., \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Lucarelli \u0026amp; Borrotti, \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2019\u003c/span\u003e).\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003eSeamless integration of ML models with blockchain infrastructure\u0026mdash;such as smart contracts and oracles\u0026mdash;offers practical pathways to transparent, decentralised, and automated trading systems. These innovations lower participation barriers and contribute to financial inclusion (Lua et al., \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Mandych et al., \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Pasupuleti, \u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e2024\u003c/span\u003e).\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003eThe identified ethical and regulatory challenges call for the development of governance frameworks and compliance mechanisms specifically tailored to AI-driven cryptocurrency trading. Addressing algorithmic biases, fraud prevention, and regulatory transparency is vital for ensuring sustainable fintech innovation (Alibašić, \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Srivastava \u0026amp; Sikroria, \u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e2024\u003c/span\u003e).\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003eReinforcement learning and multi-agent systems demonstrate practical utility in dynamic asset allocation and volatility management. Their ability to adapt to changing market regimes presents a valuable toolset for building resilient trading systems (Kang et al., \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Kumlungmak, \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Zhao et al., \u003cspan citationid=\"CR69\" class=\"CitationRef\"\u003e2024\u003c/span\u003e).\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003eThe use of sentiment analysis and alternative data sources in ML pipelines significantly enhances predictive capabilities. These findings suggest that fintech applications should integrate multimodal data to improve market forecasts and trading signal reliability (Avramelou et al., \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Belcastro et al., \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Leung et al., \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2023\u003c/span\u003e).\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003eDespite the promising outcomes, issues such as limited model generalisability, sensitivity to extreme market shocks, and unaccounted transaction costs remain critical concerns. This underscores the need for continuous empirical validation and iterative model development to ensure scalability and operational viability in real-world trading environments (Alessandretti et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Huang, \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Jabbar \u0026amp; Jalil, \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2024\u003c/span\u003e).\u003c/p\u003e\u003c/li\u003e\u003c/ul\u003e\u003c/p\u003e\u003c/div\u003e"},{"header":"4. Discussion","content":"\u003cp\u003eThis section critically interprets the reviewed literature in relation to the study\u0026rsquo;s core research questions and thematic domains. The aim is to provide a structured reflection on how the integration of machine learning (ML) into cryptocurrency trading strategies has influenced predictive performance, risk mitigation, fintech integration, and regulatory preparedness.\u003c/p\u003e\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e\u003ch2\u003e4.1 Predictive Modelling and Trading Performance\u003c/h2\u003e\u003cp\u003eThe literature strongly supports the hypothesis that ML models outperform traditional statistical methods in cryptocurrency price forecasting. In particular, long short-term memory (LSTM), gated recurrent units (GRU), random forests, and ensemble methods show consistent improvements in return predictability. These results are robust across different cryptocurrencies, time frames, and feature sets. However, their effectiveness is often context-dependent, with some models underperforming in bearish or high-volatility conditions.\u003c/p\u003e\u003cp\u003eThis indicates that while ML enhances predictive power, its generalisation remains limited by data quality and market dynamics. Nevertheless, the findings confirm that ML can significantly improve trading profitability, partially answering RQ1 regarding the impact on trading outcomes.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e\u003ch2\u003e4.2 Algorithmic Robustness and Adaptability\u003c/h2\u003e\u003cp\u003eReinforcement learning (RL) algorithms, especially those using Proximal Policy Optimisation (PPO) and Deep Q-Learning, demonstrate the ability to adapt dynamically to shifting market conditions. Multi-agent RL systems allow decentralised decision-making and portfolio optimisation in real-time environments.\u003c/p\u003e\u003cp\u003eHowever, these systems often require substantial computational resources and tuning. Despite their complexity, RL approaches provide flexibility and resilience, particularly in algorithmic trading, supporting RQ2 by showing how ML can optimise decision-making under uncertainty.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec15\" class=\"Section2\"\u003e\u003ch2\u003e4.3 Risk Management and Volatility Handling\u003c/h2\u003e\u003cp\u003eMany studies incorporate risk-adjusted metrics such as the Sharpe ratio, drawdown, and Value at Risk (VaR), indicating an evolving maturity in modelling both return and risk. Hybrid models, which combine technical indicators and sentiment analysis, provide greater stability in volatile environments.\u003c/p\u003e\u003cp\u003eYet, extreme tail risks, leverage effects, and slippage remain underexplored. This suggests that ML models still lack comprehensive frameworks for high-stress scenarios. The research partially answers RQ3, demonstrating that ML improves volatility handling, but further development is needed for robust risk management across diverse market regimes.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec16\" class=\"Section2\"\u003e\u003ch2\u003e4.4 Fintech Integration and Technical Feasibility\u003c/h2\u003e\u003cp\u003eThe integration of ML with blockchain technologies (e.g., smart contracts, decentralised applications) reveals innovative approaches for automated and transparent trading systems. Oracles and trustless verification mechanisms further support scalability.\u003c/p\u003e\u003cp\u003eHowever, interoperability, latency, and cybersecurity risks hinder real-world deployment. Many proposals remain theoretical or simulation-based. Thus, while ML shows promise in fintech innovation, RQ4 remains only partially answered, ML integration is feasible, but operational barriers constrain widespread implementation.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec17\" class=\"Section2\"\u003e\u003ch2\u003e4.5 Ethical, Regulatory, and Operational Considerations\u003c/h2\u003e\u003cp\u003eThe literature reveals growing awareness of ethical implications, including algorithmic bias, market manipulation, and investor protection. Discussions on RegTech applications and adaptive regulation frameworks highlight a path forward for responsible AI use.\u003c/p\u003e\u003cp\u003eNevertheless, few empirical studies offer actionable governance models, and most ignore legal compliance, data privacy, and explainability. This shows that RQ5 remains unanswered mainly, although ethical discourse is emerging, operational guidelines and regulatory alignment are still nascent.\u003c/p\u003e\u003c/div\u003e"},{"header":"5. Conclusion","content":"\u003cp\u003eThis systematic review examined the fintech implications of integrating machine learning (ML) algorithms into cryptocurrency trading strategies, analysing 57 peer-reviewed studies across five key thematic domains: predictive performance, algorithmic trading optimisation, risk management, technological integration, and ethical-regulatory concerns.\u003c/p\u003e\u003cp\u003eRegarding RQ1 (How does ML affect trading performance in cryptocurrency markets?), the evidence overwhelmingly supports the notion that ML algorithms, especially deep learning models such as LSTM and GRU, significantly enhance predictive accuracy and trading returns (Lu et al., \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Jabbar \u0026amp; Jalil, \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Vi\u0026eacute;itez et al., \u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). ML-based strategies consistently outperform traditional methods across various coins and timeframes, especially under favourable market regimes.\u003c/p\u003e\u003cp\u003eIn relation to RQ2 (How effective are reinforcement learning and hybrid models for portfolio optimisation?), the review found strong support for the adaptability of reinforcement learning (RL) frameworks, particularly in volatile markets. Advanced RL methods and hybrid models using actor-critic structures and multi-agent systems have demonstrated superior risk-return profiles and dynamic rebalancing capabilities (Kang et al., \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Asgari \u0026amp; Khasteh, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2021\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eFor RQ3 (To what extent can ML models manage volatility and mitigate trading risk?), most studies showed improved handling of short-term volatility and false trading signals through hybrid techniques and risk-adjusted metrics like the Sharpe Ratio and Value at Risk (Zhao et al., \u003cspan citationid=\"CR69\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Tanwar \u0026amp; Raboaca, \u003cspan citationid=\"CR63\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). However, comprehensive tail-risk modelling and robustness under black-swan events remain underexplored.\u003c/p\u003e\u003cp\u003eRQ4 (What is the feasibility of integrating ML models within real-world fintech infrastructure?) yielded mixed results. Although many studies propose ML-enhanced trading systems using blockchain, smart contracts, and decentralised automation (Lua et al., \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Mandych et al., \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2023\u003c/span\u003e), real-world deployment is rare due to scalability issues, infrastructure latency, and interoperability concerns.\u003c/p\u003e\u003cp\u003eFinally, RQ5 (How are ethical, regulatory, and governance issues addressed in ML-driven crypto trading?) remains insufficiently answered. While some recent works acknowledge the need for explainability, fairness, and investor protection (Kumari et al., \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Srivastava \u0026amp; Sikroria, \u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e2024\u003c/span\u003e), concrete legal and ethical frameworks for responsible AI deployment in crypto trading are still lacking.\u003c/p\u003e\u003cp\u003eIn conclusion, while the integration of ML into cryptocurrency trading demonstrates substantial advancements in predictive analytics, profitability, and risk control, practical and normative challenges persist. Future research should prioritise the development of robust, interpretable, and regulation-aligned ML systems to ensure both market efficiency and investor safety in the rapidly evolving fintech ecosystem.\u003c/p\u003e"},{"header":"6. Limitations and Future Research","content":"\u003cp\u003eAlthough this review provides a broad overview of the literature on machine learning applications in cryptocurrency trading, several limitations should be acknowledged. The study is restricted to peer-reviewed articles published in English between 2016 and 2025, which may have excluded relevant insights from grey literature, industry reports, or non-English academic sources. This language and publication bias may limit the global generalisability of the findings, particularly in regions where fintech innovations emerge outside the traditional academic ecosystem.\u003c/p\u003e\u003cp\u003eThe thematic scope of the review focused primarily on studies integrating machine learning algorithms into technical trading frameworks. As a result, other relevant perspectives\u0026mdash;such as behavioural finance, institutional economics, or legal analyses of algorithmic trading\u0026mdash;received limited attention. Moreover, the majority of the reviewed studies relied on historical backtesting or simulated trading environments. While these approaches offer valuable experimental insights, they do not necessarily reflect the constraints and uncertainties of real-world implementation, such as latency, transaction costs, or changing regulatory environments.\u003c/p\u003e\u003cp\u003eAnother concern is the potential for publication bias, as positive or novel outcomes are more likely to be published than negative or inconclusive findings. This may distort the perceived effectiveness of certain algorithms or modelling strategies, leading to overly optimistic interpretations of their practical viability.\u003c/p\u003e\u003cp\u003eLooking ahead, future research should prioritise empirical studies that validate machine learning models under live market conditions and quantify the impact of operational constraints on trading performance. A promising direction is the integration of explainable AI techniques and ethical design principles into cryptocurrency trading algorithms to enhance transparency, accountability, and regulatory compliance. Further exploration is also needed in merging decentralised finance infrastructures with machine learning systems, including smart contracts, blockchain oracles, and automated market-making mechanisms.\u003c/p\u003e\u003cp\u003eAnother area ripe for investigation involves enhancing model robustness under extreme market events and high-volatility scenarios. Stress-testing ML strategies during black-swan events or regulatory interventions could provide deeper insights into their reliability. Finally, as data sources become increasingly diverse and unstructured, developing models capable of fusing multimodal inputs, including on-chain data, macroeconomic indicators, and real-time sentiment, will be crucial for generating actionable trading signals.\u003c/p\u003e\u003cp\u003eBy addressing these limitations and expanding the research horizon, scholars and practitioners can better navigate the complex interplay between artificial intelligence and digital financial markets.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eFUNDING:\u003c/strong\u003e This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCOMPETING INTERESTS:\u003c/strong\u003e The authors declare no competing interests.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAUTHOR CONTRIBUTIONS:\u003c/strong\u003e P.L. conceived the study, developed the methodological framework, coordinated the research process, and contributed to the manuscript writing. J.P. performed the literature review, data analysis, and prepared the comparative tables. I.F. contributed to the interpretation of results and manuscript revision. All authors reviewed and approved the final manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDATA AVAILABILITY:\u003c/strong\u003e The study is based on publicly available academic literature and does not rely on proprietary or primary datasets. A full list of reviewed studies is included in Appendix A. Any additional materials or coding frameworks used in the thematic synthesis are available from the authors upon reasonable request.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eETHICAL APPROVAL:\u003c/strong\u003e This article does not contain any studies with human participants performed by any of the authors.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eINFORMED CONSENT:\u003c/strong\u003e This article does not contain any studies with human participants performed by any of the authors.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCONSENT TO PARTICIPATE:\u003c/strong\u003e Not applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCONSENT TO PUBLISH:\u003c/strong\u003e The authors consent to the publication of this manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCLINICAL TRIAL NUMBER:\u0026nbsp;\u003c/strong\u003eNot applicable.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eAlaminos D, Salas MB, Callej\u0026oacute;n-Gil \u0026Aacute;M. Managing extreme cryptocurrency volatility in algorithmic trading: EGARCH via genetic algorithms and neural networks. Quant Finance Econ. 2024;8(1):153\u0026ndash;209. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3934/qfe.2024007\u003c/span\u003e\u003cspan address=\"10.3934/qfe.2024007\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAlessandretti L, ElBahrawy A, Aiello LM, Baronchelli A. (2018). Anticipating cryptocurrency prices using machine learning. \u003cem\u003eComplexity\u003c/em\u003e. 2018(1), 16p. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1155/2018/8983590\u003c/span\u003e\u003cspan address=\"10.1155/2018/8983590\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAlibašić H. (2023). Developing an Ethical Framework for Responsible Artificial Intelligence (AI) and Machine Learning (ML) Applications in Cryptocurrency Trading: A Consequentialism Ethics Analysis. \u003cem\u003eFinTech\u003c/em\u003e. 2(3), pp. 430\u0026ndash;443. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/fintech2030024\u003c/span\u003e\u003cspan address=\"10.3390/fintech2030024\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAmirzadeh R, Nazari A, Thiruvady DR. Applying Artificial Intelligence in Cryptocurrency Markets: A Survey. Algorithms. 2022;15(11):428. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/a15110428\u003c/span\u003e\u003cspan address=\"10.3390/a15110428\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAmirzadeh R, Thiruvady DR, Nazari A, Ee MS. (2023). A Framework for Empowering Reinforcement Learning Agents with Causal Analysis: Enhancing Automated Cryptocurrency Trading. \u003cem\u003earXiv.Org\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.48550/arxiv.2310.09462\u003c/span\u003e\u003cspan address=\"10.48550/arxiv.2310.09462\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAsgari M, Khasteh H. (2021). Profitable Strategy Design for Trades on Cryptocurrency Markets with Machine Learning Techniques. \u003cem\u003earXiv: Trading and Market Microstructure\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.48550/arxiv.2105.06827\u003c/span\u003e\u003cspan address=\"10.48550/arxiv.2105.06827\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAvramelou L, Nousi P, Passalis N, Doropoulos S, Tefas A. (2023). Cryptosentiment: A Dataset and Baseline for Sentiment-Aware Deep Reinforcement Learning for Financial Trading. \u003cem\u003e2023 IEEE International Conference on Acoustics, Speech, and Signal Processing Workshops (ICASSPW)\u003c/em\u003e, Rhodes Island, Greece, 2023, 1\u0026ndash;5. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1109/icasspw59220.2023.10193330\u003c/span\u003e\u003cspan address=\"10.1109/icasspw59220.2023.10193330\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBelcastro L, Carbone D, Cosentino C, Marozzo F, Trunfio P. Enhancing Cryptocurrency Price Forecasting by Integrating Machine Learning with Social Media and Market Data. Algorithms. 2023;16(12):542. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/a16120542\u003c/span\u003e\u003cspan address=\"10.3390/a16120542\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBellocca GP, Attanasio G, Cagliero L, Fior J. Leveraging the momentum effect in machine learning-based cryptocurrency trading. Mach Learn Appl. 2022;8:100310. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.mlwa.2022.100310\u003c/span\u003e\u003cspan address=\"10.1016/j.mlwa.2022.100310\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBetancourt C, Chen WH. Reinforcement Learning with Self-Attention Networks for Cryptocurrency Trading. Appl Sci. 2021;11(16):7377. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/APP11167377\u003c/span\u003e\u003cspan address=\"10.3390/APP11167377\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBohr A, Memarzadeh K. (2020). Chapter 12 - Systematic Review and Meta-Analysis. In A. Bohr \u0026amp; K. Memarzadeh, editors, \u003cem\u003eArtificial Intelligence in Healthcare\u003c/em\u003e, 267\u0026ndash;285. Academic Press. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/B978-0-12-818438-7.00012-5\u003c/span\u003e\u003cspan address=\"10.1016/B978-0-12-818438-7.00012-5\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBouteska A, Abedin MZ, H\u0026aacute;jek P, Yuan K. Cryptocurrency price forecasting \u0026ndash; A comparative analysis of ensemble learning and deep learning methods. Int Rev Financial Anal. 2024;92. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.irfa.2023.103055\u003c/span\u003e\u003cspan address=\"10.1016/j.irfa.2023.103055\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBraun V, Clarke V. Using thematic analysis in psychology. Qualitative Res Psychol. 2006;3(2):77\u0026ndash;101. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1191/1478088706qp063oa\u003c/span\u003e\u003cspan address=\"10.1191/1478088706qp063oa\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBu SJ, Cho S-B. Learning Optimal Q-Function Using Deep Boltzmann Machine for Reliable Trading of Cryptocurrency. In: Yin H, Camacho D, Novais P, Tall\u0026oacute;n-Ballesteros A, editors. Intelligent Data Engineering and Automated Learning \u0026ndash; IDEAL 2018. IDEAL 2018. Lecture Notes in Computer Science(). Volume 11314. Cham: Springer; 2018. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/978-3-030-03493-1_49\u003c/span\u003e\u003cspan address=\"10.1007/978-3-030-03493-1_49\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eChen Z. (2022). A Novel Machine Learning-assisted Pairs Trading Approach for Trading Risk Reduction. \u003cem\u003e2022 IEEE 1st Global Emerging Technology Blockchain Forum: Blockchain \u0026amp;\u003c/em\u003e Beyond \u003cem\u003e(iGETblockchain\u003c/em\u003e), Irvine, CA, USA, 2022, 1\u0026ndash;6 \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1109/iGETblockchain56591.2022.10087166\u003c/span\u003e\u003cspan address=\"10.1109/iGETblockchain56591.2022.10087166\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFelizardo L, Paiva FCL, Graves CdeV, Matsumoto EY, Costa AHR, Del-Moral-Hernandez E, Brandimarte P. Outperforming algorithmic trading reinforcement learning systems: A supervised approach to the cryptocurrency market. Expert Syst Appl. 2022;202. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.eswa.2022.117259\u003c/span\u003e\u003cspan address=\"10.1016/j.eswa.2022.117259\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFeng W, Wang Y, Zhang Z. Can cryptocurrencies be a safe haven: a tail risk perspective analysis. Appl Econ. 2019;51(47):5091\u0026ndash;109. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/00036846.2019.1606405\u003c/span\u003e\u003cspan address=\"10.1080/00036846.2019.1606405\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFischer TG, Krauss C, Deinert A. Statistical Arbitrage in Cryptocurrency Markets. J Risk Financial Manag. 2019;12(1):31. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/JRFM12010031\u003c/span\u003e\u003cspan address=\"10.3390/JRFM12010031\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHafid A, Rahouti M, Kong L, Ebrahim M, Serhani MA. Predicting Bitcoin Market Trends with Enhanced Technical Indicator Integration and Classification Models. arXiv: Mach Learn. 2024. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.48550/arxiv.2410.06935\u003c/span\u003e\u003cspan address=\"10.48550/arxiv.2410.06935\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHou C. AI Technology\u0026rsquo;s Application and Impact in the Secondary Market of Virtual Currencies. J Appl Econ Policy Stud. 2025;16. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.54254/2977-5701/2025.20565\u003c/span\u003e\u003cspan address=\"10.54254/2977-5701/2025.20565\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHuang S. Enhancing investment strategies through machine learning: A comprehensive analysis across market sectors. Appl Comput Eng. 2024;104. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.54254/2755-2721/104/20240909\u003c/span\u003e\u003cspan address=\"10.54254/2755-2721/104/20240909\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eJabbar A, Jalil SQ. (2024). A Comprehensive Analysis of Machine Learning Models for Algorithmic Trading of Bitcoin. \u003cem\u003earXiv: Trading and Market Microstructure\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.48550/arxiv.2407.18334\u003c/span\u003e\u003cspan address=\"10.48550/arxiv.2407.18334\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eJi Q, Zhang D, Zhao Y. Searching for safe-haven assets during the COVID-19 pandemic. Int Rev Financial Anal. 2019;71:101526. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.irfa.2020.101526\u003c/span\u003e\u003cspan address=\"10.1016/j.irfa.2020.101526\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKang C, Woo J-S, Hong JW. (2024). A2C Reinforcement Learning for Cryptocurrency Trading and Asset Management. \u003cem\u003e2024 IEEE International Conference on Blockchain and Cryptocurrency (ICBC)\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1109/icbc59979.2024.10634413\u003c/span\u003e\u003cspan address=\"10.1109/icbc59979.2024.10634413\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKoker TE, Koutmos D. Cryptocurrency Trading Using Machine Learning. J Risk Financial Manag. 2020;13(8). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/JRFM13080178\u003c/span\u003e\u003cspan address=\"10.3390/JRFM13080178\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKouloumpris E, Vlahavas I. Combining deep reinforcement learning with technical analysis and trend monitoring on cryptocurrency markets. Neural Comput Appl. 2023;35:21445\u0026ndash;62. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s00521-023-08516-x\u003c/span\u003e\u003cspan address=\"10.1007/s00521-023-08516-x\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKristjanpoller W, Minutolo M. (2018). Forecasting volatility of cryptocurrencies using GARCH models. \u003cem\u003eApplied Sciences\u003c/em\u003e, 8(10), 1987. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/app8101987\u003c/span\u003e\u003cspan address=\"10.3390/app8101987\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKumari M, Sinha R, Chakrabarty P, Bhardwaj YK, Priya SA, Gupta S. (2024). Enhancing Digital Currency Pricing with Machine Learning Models. \u003cem\u003e2024 15th International Conference on Computing Communication and Networking\u003c/em\u003e Technologies \u003cem\u003e(ICCCNT)\u003c/em\u003e, Kamand, India, 1\u0026ndash;6. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1109/icccnt61001.2024.10725156\u003c/span\u003e\u003cspan address=\"10.1109/icccnt61001.2024.10725156\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKumlungmak K. Multi-agent deep reinforcement learning for cryptocurrency trading. Chula DigiVerse. 2024. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.58837/chula.the.2022.95\u003c/span\u003e\u003cspan address=\"10.58837/chula.the.2022.95\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLahmiri S, Bekiros S. Cryptocurrency forecasting with deep learning chaotic neural networks. Chaos Solitons Fractals. 2019;118:35\u0026ndash;40. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.chaos.2018.11.014\u003c/span\u003e\u003cspan address=\"10.1016/j.chaos.2018.11.014\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLam KY, Makarov I. (2023). Financial Trading via Reinforcement Learning and Convolutional Neural Networks. \u003cem\u003e2023 IEEE 21st Jubilee International Symposium on Intelligent Systems and Informatics (SISY).\u003c/em\u003e \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1109/sisy60376.2023.10417967\u003c/span\u003e\u003cspan address=\"10.1109/sisy60376.2023.10417967\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLeung MF, Chan L, Hung W-C, Tsoi S-F, Lam C, Cheng Y-H. (2023). An Intelligent System for Trading Signal of Cryptocurrency Based on Market Tweets Sentiments. \u003cem\u003eFinTech\u003c/em\u003e. 2(1), 153\u0026ndash;169. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/fintech2010011\u003c/span\u003e\u003cspan address=\"10.3390/fintech2010011\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLi X, Liu Y, Liu Z. Cryptocurrency return prediction: A machine learning analysis. Social Sci Res Netw. 2024. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.2139/ssrn.4703167\u003c/span\u003e\u003cspan address=\"10.2139/ssrn.4703167\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLu J, Setthawong R, Setthawong P. (2024). A Framework Utilizing Genetic Algorithm Wrapped Neural Network Model in Discovering Effective Technical Indicators for the Cryptocurrency Market. 2024 8th International Conference on Information Technology (InCIT). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1109/incit63192.2024.10810651\u003c/span\u003e\u003cspan address=\"10.1109/incit63192.2024.10810651\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLua ZZ, Seow CK, Chan RCB, Cai Y, Cao Q. Automated Bitcoin Trading dApp Using Price Prediction from a Deep Learning Model. Risks. 2025;13(1):17. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/risks13010017\u003c/span\u003e\u003cspan address=\"10.3390/risks13010017\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLucarelli G, Borrotti M. A Deep Reinforcement Learning Approach for Automated Cryptocurrency Trading. In: MacIntyre J, Maglogiannis I, Iliadis L, Pimenidis E, editors. Artificial Intelligence Applications and Innovations. \u003cem\u003eAIAI 2019. IFIP Advances in Information and Communication Technology\u003c/em\u003e. Volume 559. Cham: Springer; 2019. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/978-3-030-19823-7_20\u003c/span\u003e\u003cspan address=\"10.1007/978-3-030-19823-7_20\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLucarelli G, Borrotti M. A deep Q-learning portfolio management framework for the cryptocurrency market. Neural Comput Applic. 2020;32:17229\u0026ndash;44. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s00521-020-05359-8\u003c/span\u003e\u003cspan address=\"10.1007/s00521-020-05359-8\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMai F, Shan Z, Bai Q, Wang X, Robert, Chiang RHL. How does social media impact Bitcoin value? A test of the silent majority hypothesis. J Manage Inform Syst. 2018;35(1):19\u0026ndash;52. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/07421222.2018.1440774\u003c/span\u003e\u003cspan address=\"10.1080/07421222.2018.1440774\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMajidi N, Shamsi M, Marvasti F. Algorithmic Trading Using Continuous Action Space Deep Reinforcement Learning. arXiv: Mach Learn. 2022. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.48550/arXiv.2210.03469\u003c/span\u003e\u003cspan address=\"10.48550/arXiv.2210.03469\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMandych O, Staverska T, Maliy O. Integration of artificial intelligence into the blockchain and cryptocurrency market. Model Dev Economic Syst. 2023;2023(4). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.31891/mdes/2023-10-8\u003c/span\u003e\u003cspan address=\"10.31891/mdes/2023-10-8\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMcNally S, Roche J, Caton S. (2018). Predicting the price of Bitcoin using machine learning. In \u003cem\u003e2018 26th Euromicro International Conference on Parallel, Distributed and Network-based Processing (PDP)\u003c/em\u003e (pp. 339\u0026ndash;343). IEEE. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1109/PDP2018.2018.00060\u003c/span\u003e\u003cspan address=\"10.1109/PDP2018.2018.00060\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMikhaiylova SS, Sabirova SA. Development of a cryptocurrency trading strategy using machine learning methods. Comput Nanatechnol. 2024;11(2):11\u0026ndash;21. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.33693/2313-223x-2024-11-2-11-21\u003c/span\u003e\u003cspan address=\"10.33693/2313-223x-2024-11-2-11-21\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eNakamoto S. (2008). Bitcoin: A peer-to-peer electronic cash system. Retrieved from \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://bitcoin.org/bitcoin.pdf\u003c/span\u003e\u003cspan address=\"https://bitcoin.org/bitcoin.pdf\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eOkasov\u0026aacute; K, Košť\u0026aacute;l K. (2024). Using Machine Learning for Predicting Arbitrage Occurrences in Cryptocurrency Exchanges. \u003cem\u003e2024 IEEE International Conference on Blockchain and\u003c/em\u003e Cryptocurrency \u003cem\u003e(ICBC)\u003c/em\u003e, Dublin, Ireland, 2024, 1\u0026ndash;7. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1109/icbc59979.2024.10634339\u003c/span\u003e\u003cspan address=\"10.1109/icbc59979.2024.10634339\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eOtabek S, Choi J. Multi-level deep Q-networks for Bitcoin trading strategies. Sci Rep. 2024;14:771. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41598-024-51408-w\u003c/span\u003e\u003cspan address=\"10.1038/s41598-024-51408-w\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePage MJ, McKenzie JE, Bossuyt PM, Boutron I, Hoffmann TC, Mulrow CD, Moher D. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ. 2021;372:n71. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1136/bmj.n71\u003c/span\u003e\u003cspan address=\"10.1136/bmj.n71\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePang Y, Sundararaj G, Ren J. (2019). Cryptocurrency Price Prediction using Time Series and Social Sentiment Data. \u003cem\u003eBDCAT'19: Proceedings of the 6th IEEE/ACM International Conference on Big Data Computing, Applications and Technologies\u003c/em\u003e. 35\u0026ndash;41. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1145/3365109.3368785\u003c/span\u003e\u003cspan address=\"10.1145/3365109.3368785\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePark J, Seo Y-S. A Deep Learning-Based Action Recommendation Model for Cryptocurrency Profit Maximization. Electronics. 2022;11(9):1466. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/electronics11091466\u003c/span\u003e\u003cspan address=\"10.3390/electronics11091466\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePasupuleti MK. AI and the Future of Finance: Transforming Cryptocurrency, FinTech, and Economic Outlooks. Int J Acad Industrial Res Innovations(IJAIRI). 2024;4(4). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.62311/nesx/97883\u003c/span\u003e\u003cspan address=\"10.62311/nesx/97883\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePeng A, Ang SL, Lim CY. Automated Cryptocurrency Trading Bot Implementing DRL. Pertanika J Sci Technol. 2022;30(4). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.47836/pjst.30.4.22\u003c/span\u003e\u003cspan address=\"10.47836/pjst.30.4.22\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eQing Y, Kong Y, Sun J, Lin J. (2022a). Cryptocurrency Trading Strategies and Profitability Based on Deep Learning Algorithms. \u003cem\u003e4th International Conference on Applied Machine Learning (ICAML).\u003c/em\u003e \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1109/ICAML57167.2022.00079\u003c/span\u003e\u003cspan address=\"10.1109/ICAML57167.2022.00079\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eQing Y, Sun J, Kong Y, Lin J. (2022b). Fundamental Multi-factor Deep-learning Strategy For Cryptocurrency Trading. \u003cem\u003eIEEE 20th International Conference on Industrial Informatics (INDIN)\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1109/INDIN51773.2022.9976116\u003c/span\u003e\u003cspan address=\"10.1109/INDIN51773.2022.9976116\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eRamkumar G. (2021). Cryptocurrency Portfolio Construction Using Machine Learning Models. In: Jajuga, K., Locarek-Junge, H., Orlowski, L.T., Staehr, K, editors \u003cem\u003eContemporary Trends and Challenges in Finance. Springer Proceedings in Business and Economics\u003c/em\u003e. Springer, Cham. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/978-3-030-73667-5_7\u003c/span\u003e\u003cspan address=\"10.1007/978-3-030-73667-5_7\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSanju T, Liyanage H, Bandara K, Kandakkulama D, Silva DD, Perera J. Stock-Crypto-App \u0026ndash; Recommendation System for Stock and Cryptocurrency Market Using Cutting Edge Machine Learning Technology. Int Res J Innovations Eng Technol. 2023;7(9):110\u0026ndash;7. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.47001/irjiet/2023.709012\u003c/span\u003e\u003cspan address=\"10.47001/irjiet/2023.709012\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSattarov O, Muminov A, Lee CW, Kang HK, Oh R, Ahn J, Oh HJ, Jeon HS. Recommending cryptocurrency trading points with deep reinforcement learning approach. Appl Sci. 2020;10(4):1506. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/APP10041506\u003c/span\u003e\u003cspan address=\"10.3390/APP10041506\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSebasti\u0026atilde;o H, Godinho P. Forecasting and trading cryptocurrencies with machine learning under changing market conditions. Financial Innov. 2021;7(3). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/S40854-020-00217-X\u003c/span\u003e\u003cspan address=\"10.1186/S40854-020-00217-X\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSingh J, Thulasiram RK, Thavaneswaran A. (2022). LSTM based Algorithmic Trading model for Bitcoin. \u003cem\u003eIEEE Symposium Series on Computational Intelligence\u003c/em\u003e. 344\u0026ndash;351 \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1109/SSCI51031.2022.10022021\u003c/span\u003e\u003cspan address=\"10.1109/SSCI51031.2022.10022021\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eŚlepaczuk R, Zenkova M. Robustness of Support Vector Machines in Algorithmic Trading on Cryptocurrency Market. Cent Eur Economic J. 2019;5(52). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1515/CEEJ-2018-0022\u003c/span\u003e\u003cspan address=\"10.1515/CEEJ-2018-0022\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSong D. Machine Learning for Price Prediction and Risk-Adjusted Portfolio Optimization in Cryptocurrencies. Mach Learn Model Techniques Financial Data Sci. 2025. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.4018/979-8-3693-8186-1.ch013\u003c/span\u003e\u003cspan address=\"10.4018/979-8-3693-8186-1.ch013\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSrivastava V, Sikroria R. Ai and algorithmic trading: a study on predictive accuracy and market efficiency in fintech applications. ShodhKosh J Visual Perform Arts. 2024;5(1). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.29121/shodhkosh.v5.i1.2024.2797\u003c/span\u003e\u003cspan address=\"10.29121/shodhkosh.v5.i1.2024.2797\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSruthi PJ, Shahithabanu M. AI Applications in Analysing and Predicting Cryptocurrency Market. Recent Trends Manage Commer. 2024;5(2):42\u0026ndash;6. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.46632/rmc/5/2/8\u003c/span\u003e\u003cspan address=\"10.46632/rmc/5/2/8\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSu PP. Adaptive Hybrid Optimized LSTM Models: A Novel Computational Framework for Algorithmic Financial Trading System. J Industrial Eng Appl Sci (JIEAS). 2024;2(6). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.70393/6a69656173.323338\u003c/span\u003e\u003cspan address=\"10.70393/6a69656173.323338\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eTanwar S, Raboaca MS. Stochastic Neural Networks-Based Algorithmic Trading for the Cryptocurrency Market. Mathematics. 2022;10(9):1456. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/math10091456\u003c/span\u003e\u003cspan address=\"10.3390/math10091456\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eTapscott D, Tapscott A. Blockchain revolution: How the technology behind bitcoin is changing money, business, and the world. Penguin; 2016.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eToleti J, Singh S, Kumar RP. (2024). Developing Robust Crypto-Currency Trading Strategies: A Fusion of Machine Learning and Hidden Markov Modelling for Enhanced Performance. \u003cem\u003eIEEE International Conference on Electronics, Computing and Communication Technologies (CONECCT)\u003c/em\u003e, Bangalore, India, 1\u0026ndash;6. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1109/conecct62155.2024.10677046\u003c/span\u003e\u003cspan address=\"10.1109/conecct62155.2024.10677046\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eVi\u0026eacute;itez A, Santos M, Naranjo R. Machine learning Ethereum cryptocurrency prediction and knowledge-based investment strategies. Knowl Based Syst. 2024;299:112088. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.knosys.2024.112088\u003c/span\u003e\u003cspan address=\"10.1016/j.knosys.2024.112088\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eYang Y, Zhang Z, Wang L. Cryptocurrency price prediction using deep learning and attention mechanism. PeerJ Comput Sci. 2020;6:e279. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.7717/peerj-cs.279\u003c/span\u003e\u003cspan address=\"10.7717/peerj-cs.279\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eŻbikowski K. Application of Machine Learning Algorithms for Bitcoin Automated Trading. In: Ryżko D, Gawrysiak P, Kryszkiewicz M, Rybiński H, editors. Machine Intelligence and Big Data in Industry. Studies in Big Data. 19 ed. Cham: Springer; 2016. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/978-3-319-30315-4_14\u003c/span\u003e\u003cspan address=\"10.1007/978-3-319-30315-4_14\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhao F, Zhang M, Zhou S, Lou Q. Application of Deep Reinforcement Learning for Cryptocurrency Market Trend Forecasting and Risk Management. J Industrial Eng Appl Sci (JIEAS). 2024;2(5). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.5281/zenodo.13854906\u003c/span\u003e\u003cspan address=\"10.5281/zenodo.13854906\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZwitter A, Boisse-Despiaux M, Posada A. Blockchain for humanitarian action and development aid. J Int Humanitarian Action. 2020;5:16. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s41018-020-00072-3\u003c/span\u003e\u003cspan address=\"10.1186/s41018-020-00072-3\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"discover-artificial-intelligence","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"diai","sideBox":"Learn more about [Discover Artificial Intelligence](https://www.springer.com/44163)","snPcode":"","submissionUrl":"","title":"Discover Artificial Intelligence","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Discover Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Cryptocurrency trading, Machine learning, Reinforcement learning, Fintech, Risk management, Systematic review","lastPublishedDoi":"10.21203/rs.3.rs-7411123/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7411123/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eThis review synthesizes research on fintech implications of integrating machine learning algorithms into cryptocurrency trading strategies to address the fragmented understanding of their impact on trading efficacy, risk management, and financial innovation. The review aimed to evaluate current knowledge on machine learning applications, benchmark algorithmic trading performance, identify risk mitigation techniques, compare algorithm effectiveness, and examine regulatory and ethical considerations. A systematic analysis of diverse methodologies, including supervised, reinforcement, and hybrid learning models across global computational finance and AI literature, was conducted. Findings indicate that deep learning and ensemble methods significantly enhance predictive accuracy and trading profitability under volatile market conditions, while reinforcement learning frameworks improve dynamic portfolio optimization and risk-adjusted returns. Risk management benefits arise from integrating technical indicators and reward-based safety mechanisms, though universal frameworks remain lacking. Fintech integration advances through blockchain-enabled transparency and automation, yet practical deployment faces scalability and interoperability challenges. Ethical and regulatory discourse is nascent, underscoring the need for responsible AI frameworks to ensure market integrity and investor protection. These findings collectively demonstrate that machine learning substantially transforms cryptocurrency trading strategies, offering enhanced performance and risk control within evolving fintech infrastructures, while highlighting critical gaps in regulatory compliance and ethical governance that warrant focused future research.\u003c/p\u003e","manuscriptTitle":"Machine Learning Integration in Cryptocurrency Trading: A Systematic Review of Fintech Implications","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-10-27 11:32:17","doi":"10.21203/rs.3.rs-7411123/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-11-05T06:20:16+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-10-22T05:40:24+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-10-22T00:21:15+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-10-20T13:57:08+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"237155295348007027818306878781336053308","date":"2025-10-15T09:51:34+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"326590457710961259623834491474205968877","date":"2025-10-14T17:04:49+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"69059607279419262224486108647828484837","date":"2025-10-14T02:38:36+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"48447468338514402535132778964672679065","date":"2025-10-13T14:17:44+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-10-13T08:42:15+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2025-09-30T19:26:15+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-08-28T10:43:30+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-08-28T10:42:49+00:00","index":"","fulltext":""},{"type":"submitted","content":"Discover Artificial Intelligence","date":"2025-08-19T18:14:39+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"discover-artificial-intelligence","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"diai","sideBox":"Learn more about [Discover Artificial Intelligence](https://www.springer.com/44163)","snPcode":"","submissionUrl":"","title":"Discover Artificial Intelligence","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Discover Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"308a756a-e479-49c0-ae53-e293894fd7c2","owner":[],"postedDate":"October 27th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[],"tags":[],"updatedAt":"2026-02-16T16:10:00+00:00","versionOfRecord":{"articleIdentity":"rs-7411123","link":"https://doi.org/10.1007/s44163-025-00785-w","journal":{"identity":"discover-artificial-intelligence","isVorOnly":false,"title":"Discover Artificial Intelligence"},"publishedOn":"2026-02-13 15:58:09","publishedOnDateReadable":"February 13th, 2026"},"versionCreatedAt":"2025-10-27 11:32:17","video":"","vorDoi":"10.1007/s44163-025-00785-w","vorDoiUrl":"https://doi.org/10.1007/s44163-025-00785-w","workflowStages":[]},"version":"v1","identity":"rs-7411123","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7411123","identity":"rs-7411123","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}