Financial ratios as indicators in bankruptcy prediction: A comparative analysis of statistical and machine learning models

Financial ratios as indicators in bankruptcy prediction: A comparative analysis of statistical and machine learning models | 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 Research Article Financial ratios as indicators in bankruptcy prediction: A comparative analysis of statistical and machine learning models Thu Hien Bui, Thi Thuy Duong Truong, Thi Phuong Thao Tran This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-5693739/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract This paper investigates the optimal approach for predicting corporate bankruptcy risk in Vietnam, utilizing a unique dataset of listed firms from 2010 to 2021 based on financial ratios. The results confirm that machine learning models significantly outperform traditional logistic regression, with XGBoost and Random Forest demonstrating superior predictive power compared to K-Nearest Neighbor and logistic regression across both one-year and two-year forecast horizons. The study also contributes methodologically by incorporating additional evaluation metrics including F1 Score, AUC-ROC, Brier Score, and Log-loss to assess classification and probability prediction performance more comprehensively. Model performance is further enhanced through hyperparameter tuning, class imbalance adjustment, and grid search optimization. Empirical findings highlight the importance of all six financial indicator groups, with specific ratios such as Total Assets Turnover, Current Assets to Sales, Current Assets to Total Assets, Equity to Total Assets, and Total Equity Growth playing a critical role in predicting corporate failure. These indicators emphasize the importance of liquidity management, asset efficiency, and equity growth in determining a firm’s financial resilience. Overall, this study not only enhances forecasting accuracy through advanced modeling but also provides valuable insights for stakeholders, particularly financial institutions, investors, and corporate managers supporting more informed decision-making and proactive risk management in Vietnam’s dynamic and evolving business environment. JEL Codes: G33, G34, M10. bankruptcy prediction financial ratios machine learning logistic regression emerging country Figures Figure 1 Figure 2 1. Introduction Bankruptcy prediction is critical to identifying potential risks of failure for the firms and other stakeholders such as investors, financial institutions, and governments (Zywicki, 2008; Liang et al., 2014). The capital markets have been growing sharply these days, and continually breaking previous peaks requires serious concentration on preventing and mitigating financial frauds and crises. Although the growth of the capital market presents many opportunities, it is also accompanied by risks for the firms and their stakeholders. Therefore, many researchers have employed different techniques to accurately estimate the likelihood of firm failure. The models are categorized into two types, namely statistical and machine learning techniques. Nevertheless, as a consequence of the market's growth in scale and complexity, predictions of statistical models are questioned. In the meantime, machine learning was utilized for better data processing and superior predictive model construction. Machine learning models are proven to have solved predictive errors recognized in traditional methods; hence, they are more favored in recent studies on bankruptcy prediction. Literature relies primarily on financial ratios computed from financial statements to forecast bankruptcy and financial distress. Besides, few studies add firms' credit attributes and characteristics as inputs for predictive models. However, financial ratios, which are complex information, are believed to be more objective, stable inputs for prediction. According to Liang et al. (2014), financial ratios can be classified into seven categories: solvency, profitability, cash flow ratios, capital structure ratios, turnover ratios, and growth. However, selecting these ratios for predictive models is still debated. According to Bellovary et al. ( 2007 ), there is a probability that a model with fewer ratios generates a more precise forecast compared to those that capture a lot. Few studies also demonstrate differences in the predictive power of distinct financial groups (Smith and Alvarez, 2021 ; Son et al., 2019 ). However, only a few findings generally cover the relevance of financial ratio groups in predicting bankruptcy. The development of intelligent methods and new approaches raises the performance of failure risk forecasting models (Duénez-Guzmán and Vose, 2013 ). This is very important due to the selection of bankruptcy-predicting models depending on the firm's characteristics and data availability. The above gaps bring us to set up two separate challenges for predicting bankruptcy: The first task is to forecast the company's financial distress status using financial ratios two and one year before the event. In this task, we employ four statistical and machine learning models. Regarding statistical models, we use Linear Regression (LR) while Random Forest (RF), Extreme Gradient Boosting (XGB), and K-Nearest Neighbor (KNN) are used as machine learning models. The accuracy rate of each model in two periods will be computed to conclude which models are the most superior for the Vietnamese market and which other emerging markets share the same characteristics. The study identifies XGB (Extreme Gradient Boosting) as the most influential predictive model, followed by RF (Random Forest), and LR (Logistic Regression) as the least efficient. This highlights the effectiveness of machine learning techniques like XGB and RF in predicting bankruptcy in emerging markets, which might have been underexplored previously in the context of Vietnam. In addition, the study evaluates the accuracy of bankruptcy prediction models over different time horizons and finds that XGB and RF remain highly accurate up to two years ahead of the possibility of bankruptcy. Moreover, novel contributions of our study include the use of F1 Score, AUC-ROC, Brier Score, and Log-loss to enhance probability evaluation, alongside hyperparameter tuning, class imbalance handling, and grid search optimization to improve model performance. Finally, results confirm the relevance of all six financial indicator groups, with key ratios such as Total Assets Turnover, Current Assets to Sales, Current Assets to Total Assets, Equity to Total Assets and Total Equity Growth play a crucial role. These variables emphasize that a firm’s ability to manage its liquidity, efficiently utilize assets, and sustain capital growth plays a decisive role in determining its financial resilience. The integration of such indicators into advanced predictive models like XGBoost not only enhances predictive accuracy but also provides practical guidance for stakeholders, particularly financial institutions, investors, and corporate managers in developing early warning systems and strategic interventions to mitigate bankruptcy risk in the Vietnamese business context. This study is divided into six sections. This beginning part is an introduction, the first section, followed by the second section, where prior studies are reviewed to uncover research gaps and facilitate the construction of this study's research direction. Subsequently, the third section, Methodology, will provide details on research models. The fourth section clarifies the data used in the study. The fifth section presents findings before the general summary in the final quarter. 2. Literature review Bankruptcy refers to a firm's inability to accomplish its financial obligations, accompanied by a declaration of default by the court ruling. The event of bankruptcy risk or financial distress poses an excellent threat to the firm and its stakeholders. Although regulations and laws have been reinforced to ease the potential damages, the growing complexity of businesses and financial involvements, as well as the lack of a mature theoretical framework on corporate failure (Wang et al., 2014 ), voice a need for exploration on determinant factors along with highly-accurate predictive models. In financial management, credit risk is minimized by assessing firms' economic well-being, which signals the likelihood of financial distress and bankruptcy. These assessments and forecasts act as crucial inputs for the decision-making process of firms' managers, investors, creditors, and policymakers (Zywicki, 2008; Liang et al., 2014). Acknowledging the vital role of bankruptcy prediction, since the late 19th century, numerous researchers have conducted to develop and propose predictive models on bankruptcy to suggest early solutions and prevent detrimental effects. The literature mainly uses financial ratios computed from annual financial statements as critical indicators for bankruptcy (Mai et al., 2018). However, few studies also employ financial or credit attributes such as credit scores, company characteristics, etc., as inputs for their research models. Regarding research techniques, they are categorized into two types: traditional models and machine learning ones. Specifically, traditional models focus on statistical methods, among which the multivariate discriminant analysis model (MDA) and logistic regression model (LR) are the most popular and widely used methods in both academic studies and practice (Ohson, 1980; Ohson et al., 2012). Machine learning models are also termed artificial intelligence models in which data are processed and used for training and testing models, ensuring higher prediction validity compared to statistical methods. The early days of bankruptcy prediction officially began with the study of Beaver ( 1966 ) using a linear regression model to classify failed and non-failed firms. The author employed 30 financial ratios specified into six categories: Cash flow ratios, Net income ratios, Debt to total asset ratios, Liquid asset to total asset ratios, Liquid asset to current debt ratios, and Turnover ratios. Noticeably, Beaver ( 1966 ) examined the firms' performances five years before their failure, concluding that predictive power is reduced as time goes backward. Besides, the study also suggested that not all ratios affect bankruptcy similarly. Cash flow ratios are the most influential, while liquid asset ratios are the weakest determinants. In 1968, Altman proposed the best-known and most widely-used function to assess a firm's financial health. The study utilized a multiple discriminant analysis method (MDA) to build a process to compute a firm's Z-score based on five financial ratios in each profitability, activity, liquidity, solvency, and leverage ratio classification. The conclusions agreed with Beaver ( 1966 ) regarding the predictive accuracy over time, which led Altman ( 1968 ) to recommend that prediction should be made in less or equal to 2 years. A study by Edmister ( 1972 ) also applied MDA in exploring 19 financial ratios to predict the bankruptcy of small businesses. MDA is the most popular method utilized in the study of bankruptcy. In 1980, the logit analysis technique, particularly maximum likelihood estimation, was used by Ohlson ( 1980 ) to uncover the effects of 6 ratios on the possibility of failure. The author also suggested further discoveries on other predictors. Nonetheless, since the 1990s, traditional techniques have been proven to be outperformed by modern models (Alaka, 2018; Du Jardin, 2015 ; Kasgari et al., 2013 ; Mai et al., 2019 ; Min and Lee, 2005 ; Ohson et al., 2012; Ohson and Sharda, 1990; Wang et al., 2012 ; Wilson and Sharda, 1994 ), artificial intelligence (AI) have been introduced as an alternative for assessing firms' well-being, and predicting corporate failure. As one of the earliest studies on the use of machine learning to predict bankruptcy, Odom and Sharda ( 1990 ) made a comparison on the accuracy of MDA and neural network technique (NN) in forecasting bankruptcy using the same five ratios as Altman's (1968). It was found that neural networks were more robust and had a higher accuracy rate than MDA, which lay a foundation for the application of modern techniques. Studies by Kasgari et al. ( 2013 ) and Son et al. ( 2019 ) also investigated the performance of neural networks over the logistic regression method. The results demonstrated the superiority of machine learning methods over statistical ones. The rapid development of data mining presents many machine-learning models. Hence, studies have been conducted with several models to compare and decide on the most potent model. Besides, ensembles of various machine learning models such as Random Forest (RF) or Extreme Gradient Boosting (XGB) are also suggested (Perboli and Arabnezhad, 2021 ; Son et al., 2019 ; Wang et al., 2011) which are proven to strengthen predictive power. The table below demonstrates past studies using machine learning models with the best-performed models in bold: Table 1 Past machine learning studies on bankruptcy prediction No. Work Models Ratio groups Dataset 1 Odom and Sharda ( 1990 ) MDA, NN Altman US 2 Wilson and Sharda ( 1994 ) SNN , MDA Altman US 3 Huang et al. ( 2004 ) SVM , BPN Profitability, leverage, liquidity, actitivy, turnover ratio US, Taiwan 4 Min and Lee ( 2005 ) SVM , MDA, LR, SNN Profitability, leverage, liquidity, actitivy, turnover ratio Korea 5 Tsai ( 2008 ) ANN, MLP , SVM US, Australia, Taiwan, Korea, German 6 Tseng and Hu ( 2010 ) LR, QIL, MLP, RBFN Management inefficiency, capital structure, insolvency, adverse economic effects, income volatility, UK 7 Chaudhuri and De ( 2011 ) SVM, FSVM Profitability, leverage, liquidity, actitivy, turnover ratio US 8 Ohson et al. (2012) ANN, DT , SVM, LR Altman US 9 Wang et al. ( 2012 ) RF , LR, MDA, MLP, RBFN Attributes not ratios Australia, German 10 Kasgari et al. ( 2013 ) ANN , LR Liquidity, profitability, leverage Iran 11 Tsai et al. ( 2014 ) SVM, MLP, DT Credit data sets Australia, German, Japan 12 Du Jardin ( 2015 ) MDA, LR, MLP Liquidity, solvency, profitability, financial structure, activity, turnover France 13 Liang et al. ( 2016 ) SVM , KNN, CART, MLP, NB Solvency, capital structure, profitability, turnover ratios, cash flow ratios, growth, corporate governance indicators Taiwan 14 Mai et al. ( 2019 ) LR, SVM, RF Liquidity, profitability, leverage US 15 Son et al. ( 2019 ) RF , DT, XGB, KNN Cash flow ratios, growth, leverage, liquidity, activity, profitability, solvency Korea 16 Muslim and Dasril ( 2021 ) KNN, DT, SVM, RF, XGB Poland 17 Narvekar and Guha ( 2021 ) RF, SVM, XGB US 18 Perboli and Arabnezhad ( 2021 ) RF, XGB , LR, NN Profitability, actitivy, leverage, liquidity Italy 19 Smith and Alvarez ( 2021 ) LR, SNN, RF , SVM, XGB, Profitability, actitivy, leverage, liquidity, turnover Spain 20 Shetty et al. ( 2022 ) NN, SVM , XGB Liquidity, solvency, turnover Belgium 21 Bragoli et al. ( 2022 ) LR, NN, RF, XGB Profitability, liquidity, growth Italy * ANN: Artificial Neutral Networks; BPN: Back-Propagation Neural Networks; CART: Classification And Regression Tree; CBR: Case-Based Reasoning; DT: Decision Trees; IF: Isolation Forest; KNN: K-Nearest Neighbor; LDA: Linear Discriminant Analysis; LR: Logistic Regression; LSAD: Least-Squares Anomaly Detection; MLP: Multi-Layer Perceptron; NB: Naive Bayes; NN: Neural Networks; OCSVM: One Class SVM; QIL: Quadratic Interval Logit Model; RBFN: Radial Basis Function Network; SAT: Single Attribute Test; SNN: Shallow Neutral Network; SVM: Support Vector Machines; XGB: Extreme Gradient Boosting. Source: By the authors Table 2 Predictive accuracy by model No. Model Lowest accuracy Highest accuracy Study with highest accuracy 1 RF 63.90% 99.68% Narvekar and Guha ( 2021 ) 2 XGB 81.00% 98.70% Narvekar and Guha ( 2021 ) 3 KNN 60.20% 95.60% Muslim and Dasril ( 2021 ) 4 DT 73.33% 94.80% Muslim and Dasril ( 2021 ) 5 SVM 66.10% 90.10% Narvekar and Guha ( 2021 ) 6 MLP 61.40% 93.75% Tseng and Hu ( 2010 ) 7 NN 71.00% 98.00% Wilson and Sharda ( 1994 ) 8 LR 49.10% 90.80% Huang et al. ( 2004 ) 9 MDA 37.50% 95.60% Wilson and Sharda ( 1994 ) Source: By the authors In general, it is noticed that the later the studies are, the more popular the ensemble techniques are, of which superiority is evidenced (Mai et al., 2019 ; Perboli and Arabnezhad, 2021 ). Studies on bankruptcy prediction frequently mention Type I and Type II errors. These are errors regarding the misclassification of the firms’ financial situation. Type I errors wrongly forecast bankrupt firms as non-bankrupt, while Type II errors happen when non-bankrupt firms are misclassified as bankrupt. Type I is argued to be the more costly of the two types of error because it poses more damages regarding a firm’s reputation, loss of its shareholders, and potential court costs (Bellovary et al., 2007 ). Hence, although studies try to handle both types of errors, models reducing the frequency of Type I errors are more favoured. Among the models used, RF, XGB, and KNN are the most popular and accurate methods used to construct predictive models on firm failure. These models are known to have resolved the issues of data dimensionality, scalability, and accuracy of classifications which confronted by previous models (Belgiu and Dragut, 2016; Bentejac et al., 2021; Biau, 2012 ; Breiman, 2001 ; Chen and Guestrin, 2016 ; Ramraj et al., 2016 ; Zhang, 2017 ). Regarding the data captured by the models, the majority of researchers use financial ratios classified as listed in Table 1 . In short, ratios uncovering firms’ liquidity, profitability, cash flow, capital structure, turnover, and growth are all studied. Nevertheless, only several studies comment on the differences in the predictive power of specific financial categories. A study by Son et al. ( 2019 ) suggests that among six groups, liquidity, solvency, and growth ratios are more relevant to bankruptcy forecast, while leverage and capital structure ratios are proven to be closely related to predictive performance, according to Smith and Alvarez ( 2021 ). In contrast, Liang et al. ( 2016 ) conclude the crucial roles of solvency and profitability ratios in the prediction. In addition, machine learning models provide quite high results for the bankruptcy prediction problem (Table 2 , see for example). This study extends the bankruptcy prediction literature in three ways. Firstly, the most popular and accurate machine learning models, inlcuding RF, XGB, and KNN will be utilized in this study. Moreover, a statistical model, namely LR, will also be run to compare and decide which model is the most suitable for the Vietnamese market and other similar emerging markets. Secondly, we also input 62 financial ratios classified into six groups to strengthen the predictive power of the models. Lastly, the influences of ratio groups on forecasting firm failure are examined to determine which groups are more correlated with bankruptcy prediction. 3. Methodology This study uses supervised machine learning models, including Random Forest (RF), Extreme Gradient Boosting (XGB), and K-Nearest Neighbor (KNN), to construct a function to make predictions on the health of firms. 3.1 Random Forest Random Forest is an ensemble bagging tree-based learning algorithm introduced by Breiman ( 2001 ). The mechanism of the ensemble model is to integrate different classifiers to create novel ones more efficiently and accurately. The Random Forest Classifier is constructed using decision trees from bootstrapped data, trees have been trained using randomly chosen subsets of both the training set and the feature set. The following steps perform the random forest: Let the data be n observations and p features. Step 1. From the train data, randomly bootstrapped sample set of size n. Step 2. Select randomly k variables from p features. Step 3. Create the decision trees from the collected data. In light of the k feature vector, each tree in the collection casts a vote to categorize the sample. The final outcome obtained the majority of votes. The Random Forest approach uses randomly selected training data and attribute subsets; each decision tree may be weak in classification and result in a significant degree of bias. However, the method's end product is an aggregate of multiple decision trees, the information from the trees will complement one another and provide a model with accurate prediction outcomes. The ability to deal with outliers and noise is the advantage of RF (Yeh et al., 2014 ). Besides classification and prediction, it can determine the importance of variables in the model, which enhances the model's performance and reduces the data dimension (Maione et al., 2016 ). In the financial industry, Random forest has been used to detect credit fraud (Whitrow et al., 2009 ) and predict customer churn using bank services (Xie et al., 2009 ), and bankruptcy prediction (Zoričák et al., 2020 ). 3.2 Extreme Gradient Boosting Extreme Gradient Boosting is an ensemble method called boosting, which has been widely used recently because of its dominance (Barboza et al., 2017 ). Models are continuously trained to allow each model to get better and make up for the shortcomings of the preceding one. Different boosting methods develop and aggregate weak learners differently during the sequential stacking process. XGB, proposed by Fiedman (2001), is a distributed gradient-boosting library developed to be very effective, adaptable, and portable. By integrating a convex loss function based on the difference between the expected and actual outputs and a penalty term for model complexity, XGB minimizes a regularized (L1 and L2) objective function. To create the final forecast, past trees are blended with residuals or errors from earlier trees as the training process continues. 3.3 K-Nearest Neighbor KNN is one of the most straightforward classifier algorithms depending on the distance (Zhao et al., 2009 ). The distance measures the similarity between the classified set and the training set. In the KNN algorithm, after determining the number of k nearest neighbors, the algorithm finds out the k-labeled sets that are nearest the unlabeled partner, and the partner is assigned a label by majority vote amongst its k-nearest neighbors (Hand et al., 2001 ; Zhao et al., 2009 ). The KNN algorithm is straightforward to utilize as it only depends on the popular distances. The technique is implemented without making any assumptions about how data are distributed. The methodology is memory-based, and new training data is updated immediately after each classification step. The number of k-neighbors is delicate, affecting the testing model's performance. 3.4 Logistic regression The logistic regression model is one of the traditional statistical methods that investigates the relationship between explanatory variables and dependent variables. Many studies were effectuated to forecast the health of companies (Du Jardin, 2010 ; Olson et al., 2012 ; Maalouf and Trafalis, 2011; Kim, 2011; Tsai, 2014). The dependent variable in a logistic regression model may be binary or categorical, and the independent variables may be a combination of continuous, flat, and binary variables. The model is formed as $$\:z={\beta\:}_{0}+{\beta\:}_{1}{X}_{1}+{\beta\:}_{2}{X}_{2}+...+{\beta\:}_{m}{X}_{m},\:y=\frac{{e}^{z}}{1+{e}^{z}}$$ where \(\:{X}_{1},...,{X}_{m}\) are explanatory variables, the output belongs to (0,1), and the object is classified into a positive class if the output is greater than 0.5, else a negative class. 3.5 Performance comparison measures Confusion matrix The prediction task can be conceptualized as a binary prediction, where the bankruptcy class is represented as a positive status (1) and the health class as a negative status (0). We use various measures to evaluate the effectiveness of the forecasting model. The confusion matrix is displayed in Table 3 (Son et al., 2019 ; Shrivastava et al., 2020 ). Table 3 Confusion matrix Actual class 1 0 Predicted class 1 True positive (TP) False positive (FP) 0 False negative (FN) True negative (TN) Source: By the authors The positive class is anticipated to be positive, according to the true positive (TP). The same idea is known as true negative for actual negative class (TN). False positives (FP) are data points forecasted as positive in negative actual class. When the actual value is positive to be negative forecasted, the outcome is categorized as false negative (FN). The model's performance is determined via accuracy, precision and recall criteria. They stem from the confusion matrix, which is described in Table 3 . $$\:\text{Accuracy}=\frac{TP+TN}{TP+FP+TN+FN}$$ The proper forecast ratio for positive and negative classes is extracted by accuracy, a crucial performance metric. However, the number of firms classified into two categories is imbalanced, and the percentage of correct prediction needs to be increased to evaluate the efficient model. We also use additional helpful measures that, depending on the target stakeholders, can be used to describe the performance of the models. These are the precision, and recall for each class, which are defined as follows: $$\:\text{Precision}=\frac{TP}{TP+FP},\:\text{Recall}=\frac{TP}{TP+FN}$$ Precision is the percentage of data items that are genuinely positive compared to those that the model has classified as positive (TP + FP). Recall displays the proportion of actual positive data points to those that are positive (TP + FN). In addition, to evaluate the models, we also use F1 score, AUC-ROC, log- loss and brier score. The F1 score shows the model’s performance, it balances precision and recall in classification models. AUC-ROC binary classifier evaluates the discrimination between positive and negative classes. The log-loss is used to assess a classification model's probability-based predictions. It gauges how closely the actual class labels match the predicted probability. The Brier score measures the accuracy of the predicted probability. A good model has excellent accuracy, precision, recall, F1 score, AUC-ROC and low log-loss, brier score. 3.6 Dataset To forecast the bankruptcy of firms in Vietnam, the data is selected from the primary financial sources available on Fiingroup.com from 2010 to 2021. The data is related to six financial indicator groups representing liquidity, capital structure, profitability, turnover ratio, cash flow ratios, and growth. We extracted and computed 62 financial criteria that presented as explanatory variables and the status of firms as dependent variables (failure and non-bankruptcy status firms, encoded the value one and zero, respectively). The definition of variables is described in Appendix 1 for more detail. After cleaning, processing and filtering data with missing values, the database is divided into two sets for two main tasks: one-year ahead bankruptcy prediction and two-year ahead bankruptcy prediction. The first task extracts financial variables of firms from 2010 to 2020 for the prediction of company status in the next year using the last fiscal year available. For example, for 2010, the machine learning models train and fit their models using explanatory variables of that year for prediction in 2011. The second set includes variables from 2010 to 2019 to forecast the health of a company in the next two years. For example, the descriptive statistics of the first set’s independent variables are listed in Appendix 2–7. As summarized in Appendix 2–7, most financial ratios witness a significant spread of their data. In other words, the firms studied in this work have quite different characteristics. The ensemble approach does not set strict conditions as conventional regressions do. If the variables in the model have a strong correlation, traditional regression can result in the multicollinearity problem. As a result, the prediction's accuracy is decreased. Figure 1 shows the correlation of variables related to firm liquidity. There are some high correlation variables such as the current ratio and quick ratio, current ratio and quick assets/current liability, total liabilities and total assets, and current liability to total assets. However, in intelligent methods like RF, XGB can handle the multicollinearity using decision trees. The models only use a few regressors to enhance prediction accuracy and resist multicollinearity issues (Sandri and Zuccolotto, 2008 ). Appendix 8–12 show the high correlation of variables, to name a few. In addition, there is a high correlation between all explanatory variables without affecting the effectiveness of the models. 4. Results and Discussions In order to accomplish two tasks for one-year ahead and two-year ahead prediction, each dataset is split into two data sets for the training and testing model using cross-validation methods. We repeat the number of fold for the best performance, the suitable number of k folds are 3 and 4 for one-year ahead and two-year ahead. The training sets are used to train the model, including Random Forest (RF), K-Nearest Neighbor (KNN), Extreme Gradient Boosting (XGB), and Logistic Regression (LR). Training outcomes will be exerted to test performance and prediction. The first training data set is used to train the forecasting model healthy status of firms for the next year using the previous year’s data, the health of firms in the next two years is predicted based on the second training data. All models use the same training set for binary classification concerning positive and negative classes. To optimize the performance of the models, we set hyperparameters to find learning rate, the depth of trees, minimal samples in leaf and split, the number of features for splitting, class weight to avoid imbalance. The matrixes are represented in Table 5 and Table 6 for two tasks on the test sets. Table 4 Performance of one-year ahead forecasting Model Accuracy Precision Recall F1-score AUC-ROC Log-loss Brier Score KNN 0.8834 0.8918 0.98 0.9338 0.8819 0.4059 0.0867 RF 0.9428 0.9591 0.9733 0.9662 0.9773 0.149 0.042 XGB 0.9369 0.9494 0.9757 0.9624 0.9757 0.1585 0.0463 LR 0.8465 0.972 0.841 0.902 0.9335 0.3901 0.1075 Source: By the authors Table 5 Performance of two-year ahead forecasting Model Accuracy Precision Recall F1-score AUC-ROC Log-loss Brier Score KNN 0.869 0.873 0.983 0.925 0.872 0.46 0.096 RF 0.931 0.945 0.975 0.959 0.963 0.184 0.053 XGB 0.933 0.953 0.967 0.96 0.966 0.174 0.05 LR 0.905 0.933 0.953 0.942 0.936 0.274 0.071 Source: By the authors As shown in Tables 4 and 5 , the models accurately classify the firms into two classes. Confirming findings of prior studies (Bragoli et al., 2022 ; Muslim and Dasril, 2021 ; Narvekar and Guha, 2021 ; Perboli and Arabnezhad, 2021 ), RF outperforms in comparison to RF, KNN, and XGB with accuracy, precision and recall, AUC-ROC, F1 socre of over 93 percent in one year ahead case, followed by XGB, KNN, while LR gets the lowest prediction. XGB is more efficient than RF for a two-year ahead set, however, this diference is not significant. So, RF and XGB outperformance in risk prediction. The outcomes show that it slightly decreases the predicted performance compared to two periods; the forecasting after one year is higher than two years ahead. In other words, the predictive performance is better as it comes nearer to the event of bankruptcy, which is similar to previous works (Mai et al., 2018; Zoricak et al., 2020; Narvekar and Guha, 2021 ; Smith and Alvarez, 2021 ). While high Recall indicators illustrate a low probability of bankrupt firms being omitted, Precision measures the accuracy of prediction; in other words, the higher Precision is, the more accurate the forecast is, high AUC-ROC shows the significant classification, low log-loss and brier score indicate that the predicted probabilities are close to the actual outcomes. In short, a good model is one in which both Recall, Precision, AUC-ROC, F1 score are high. The results of the firm's risk prediction demonstrate the superior performance of the models, with high accuracy and strong discriminatory power. In an overall comparison, KNN produced the least favorable prediction results, with higher log-loss and Brier score compared to the other three models. Moreover, in all models, there are no significant changes regarding the accuracy rate, which demonstrates that bankruptcy can be precisely predicted up to two years before the event of default. To uncover the most correlated financial ratio groups with bankruptcy prediction, one-year ahead forecasts are run except for one particular group among liquidity ratios, capital structure ratios, profitability ratios, turnover ratios, cash flows ratios, and growth ratios. The results are demonstrated in Table 6 below: Table 6 Performance of one-year ahead forecasting without a group of financial ratios Model Accuracy Precision Recall F1-score AUC-ROC Log-loss Brier Score Without liquidity group KNN 0.879 0.888 0.98 0.932 0.879 0.423 0.088 RF 0.941 0.959 0.971 0.965 0.975 0.151 0.043 XGB 0.944 0.962 0.971 0.967 0.979 0.145 0.041 LR 0.924 0.949 0.962 0.955 0.945 0.259 0.06 Without capital structure group KNN 0.898 0.909 0.976 0.942 0.909 0.336 0.078 RF 0.9428 0.959 0.973 0.966 0.977 0.149 0.042 XGB 0.944 0.962 0.971 0.967 0.978 0.145 0.041 LR 0.943 0.959 0.974 0.966 0.973 0.19 0.04 Without profitability group KNN 0.887 0.894 0.981 0.936 0.896 0.339 0.083 RF 0.946 0.967 0.969 0.968 0.98 0.132 0.039 XGB 0.943 0.96 0.973 0.966 0.979 0.159 0.042 LR 0.9428 0.961 0.971 0.966 0.971 0.21 0.042 Without turnover ratio group KNN 0.857 0.869 0.976 0.92 0.811 0.529 0.109 RF 0.947 0.968 0.969 0.969 0.978 0.142 0.04 XGB 0.947 0.966 0.971 0.969 0.980 0.133 0.039 LR 0.942 0.955 0.976 0.966 0.973 0.186 0.041 Without cash flow ratio group KNN 0.892 0.9 0.98 0.938 0.897 0.349 0.082 RF 0.947 0.964 0.973 0.969 0.979 0.135 0.039 XGB 0.944 9.962 0.971 0.967 0.978 0.145 0.04 LR 0.942 0.962 0.962 0.966 0.971 0.215 0.042 Without growth group KNN 0.886 0.893 0.981 0.935 0.885 0.374 0.086 RF 0.941 0.96 0.969 0.965 0.976 0.149 0.042 XGB 0.943 0.959 0.974 0.966 0.979 0.158 0.042 LR 0.926 0.953 0.959 0.956 0.953 0.239 0.055 Source: By the authors Table 6 shows the risk prediction results when excluding a group of financial indicators. This implies that risk prediction should consider all six aspects of financial indicators. The results also demonstrate that the predictive power of financial groups may vary across different economies due to their nature and characteristics. In addition, the findings show that the most suitable model to effectively predict bankruptcy risk for Vietnamese businesses is XGBoost. These results indicate that the machine learning model has an advantage in forecasting problems compared to the traditional logit model. At the same time, "boosting" algorithms like XGBoost often outperform other intelligent methods for forecasting problems. The results align with prior research such as Narvekar and Guha ( 2021 ), Perboli and Arabnezhad ( 2021 ), Bragoli et al. ( 2022 ). To explore which features have the influence of financial distress prediction of firms, the study uses XGB musesod to calculate Fscore. It reflects the total number of times that variable appears during the creation of data division nodes across the entire ensemble of trees of the XGBoost model. The Fscore of variables is shown in Appendix 13 and Fig. 2 . Looking at the Fig. 2 and Appendix 13, X40, X46, X6, X18, X60 which representative for Total Assets Turnover, Current Assets to Sales, Current Assets to Total Assets, Equity to Total Assets and Total Equity Growth have the most important roles in predicting the failure risk. The emphasis on these ratios highlights several key aspects. Firstly, Total Assets Turnover (X40) measures the efficiency of a company's use of its assets to generate sales. A higher ratio indicates better performance and efficiency, which is crucial for assessing the operational efficiency and effectiveness of asset utilization in predicting company success or failure. Secondly, Current Assets to Sales (X46) helps in understanding how well a company can support its sales with its current assets, which is important for short-term financial stability and liquidity. This is critical for assessing overall profitability and shareholder value. In addition, Current Assets to Total Assets (X6) shows the proportion of current assets in relation to total assets. It indicates how much of the firm's assets are in liquid form, which is essential for understanding the liquidity position and the ability to cover short-term obligations. Equity to Total Assets (X18) reflects the proportion of a company's assets financed by shareholders' equity rather than debt. This not only enhances financial stability but also lowers the risk of insolvency during periods of financial stress. Finally, Total Equity Growth (X60) reflexes financial heath of firms. The company has a strong ability to self-finance its operations, which reduces its dependence on external debt Therefore, it reduces financial risk. To wrap up, the variables X40, X46, X18, X6, and X60 collectively underline the importance of liquidity, turnover ratios, and growth in predicting the failure risk of Vietnamese firms. The reason due to the feature of financial system in Vietnam, which is the bank-based financial system. Therefore, companies that manage their turnover, maintain healthy profitability, and ensure sufficient liquidity are better positioned to avoid financial difficulties and sustain their operations in the long run. 5. Conclusions and limitations To forecast firm bankruptcies in Vietnam, data spanning from 2010 to 2021 was collected from primary financial sources available on Fiingroup.com. The dataset comprises six key financial indicator groups: liquidity, capital structure, profitability, turnover ratio, cash flow ratios, and growth. A total of 62 financial criteria were derived and calculated as explanatory variables, with the firms' statuses categorized as dependent variables (coded as one for failure and zero for non-bankruptcy status firms). For modeling purposes, we employed a combination of statistical and machine learning techniques. Linear Regression (LR) was utilized as the statistical model, while Random Forest (RF), Extreme Gradient Boosting (XGB), and K-Nearest Neighbor (KNN) were employed as machine learning models. The results have contributed crucially to the literature on bankruptcy prediction in an emerging market such as Vietnam. Specifically, the most suitable predictive model is chosen while the impacts of financial ratio groups on the forecast are examined and concluded. Regarding the models for predicting bankruptcy or financial distress, in other words, XGB is the most influential model, with an accurate rate of 93% in both periods. Ranking right behind XGB is RF, with over 93% accuracy, but is slightly lower. KNN is the least efficient model, which forecast 88.34% of bankrupt cases and the ability classification is 88.19%. Another finding to withdraw is that the accuracy slightly increases as it is nearer to the actual event. However, the accuracy rates are still high at 93%, which ensures strong predictability of XGB and RF up to two years ahead of the possibility of bankruptcy. This underscores the effectiveness of advanced machine learning techniques, which might have been underexplored previously. Unlike previous studies that primarily rely on accuracy, precision, and recall, this research introduces additional evaluation metrics, including F1 score, AUC-ROC to assess performance and discriminative power, and Brier Score and Log-loss to evaluate the quality of probability predictions. Furthermore, the study employs model fine-tuning through hyperparameter optimization, addresses class imbalance by adjusting class weights, and applies grid search to identify the most effective model structure. These enhancements significantly improve predictive performance. In addition, empirical results confirm the importance of all six financial indicator groups in forecasting bankruptcy, while also highlighting that no single group, when omitted, leads to a significant drop in accuracy, indicating model stability. Among the 62 financial ratios analyzed, variables such as Total Assets Turnover, Current Assets to Sales, Current Assets to Total Assets, Equity to Total Assets and Total Equity Growth emerge as the most influential in predicting financial failure. The findings hold practical value for multiple stakeholders. For example, financial institutions can better assess credit risk and allocate resources more effectively, investors gain a more accurate tool for evaluating firm sustainability and managing portfolio risk, and corporate managers can use these insights to strengthen internal financial structures and prevent insolvency. By enhancing the reliability and interpretability of bankruptcy prediction models, this study supports more informed decision-making and risk management strategies, particularly within the dynamic and evolving landscape of the Vietnamese business environment. This study acknowledges some limitations. Firstly, although the study successfully uncovers that each financial ratio category can have different impacts on the forecast of bankruptcy, models still need to be adjusted to strengthen the accuracy rate by re-selecting ratios. Secondly, in addition to financial ratios, credit attributes and firm characteristics are also evidenced to affect predictive power; however, they are not studied in this study. Therefore, several future research directions are suggested. Despite the increasing accuracy of bankruptcy predictive models, practical application demands a more optimized set of inputs. As revealed by the findings, it is noticed that the rate increases accurately, nonetheless insignificantly, in the absence of certain ratio groups. In addition, credit attributes and firm characteristics may also be determinants of prediction. Finally, incorporating time-series models (e.g., LSTM, ARIMA, Hidden Markov Models) to account for the evolving nature of financial health is certainly an interesting direction. These suggestions pave the way for further advancements in bankruptcy prediction research, facilitating more informed decision-making within the Vietnamese and broader emerging market contexts. Declarations *Ethical Approval and Consent to Participate The authors declare to follow the ethical guidelines and consent to participate. *Consent for Publication The authors give consent for the publication. *Funding Not applicable Author Contribution Author Contributions StatementConceptualization, B.T.H; methodology, T.T.T.D; validation, T.T.P.T ; writing—original draft preparation, B.T.H; writing—review and editing, T.T.T.D; visualization, T.T.P.T.AcknowledgementNot ApplicableFundingNot Applicable Acknowledgement not applicable Data Availability The data that support the findings of this study are openly available at https://doi.org/10.7910/DVN/6B91QV, Harvard Dataverse, V1. References Alaka, H. A., Oyedele, L. O., Owolabi, H. A., Kumar, V., Ajayi, S. O., Akinade, O. O., & Bilal, M. (2018). Systematic review of bankruptcy prediction models: Towards a framework for tool selection. Expert Systems with Applications , 94 , 164-184. Altman, E. I. (1968). Financial ratios, discriminant analysis and the prediction of corporate bankruptcy. The journal of finance , 23 (4), 589-609. Barboza, F., Kimura, H. & Alman, E. (2017). Machine learning models and bankruptcy prediction. Expert Systems with Applications, 83, 405-417. Beaver, W. H. (1966). Financial ratios as predictors of failure. Journal of accounting research , 71-111. Belgiu, M., & Drăguţ, L. (2016). Random forest in remote sensing: A review of applications and future directions. ISPRS journal of photogrammetry and remote sensing , 114 , 24-31. Bellovary, J. L., Giacomino, D. E., & Akers, M. D. (2007). A review of bankruptcy prediction studies: 1930 to present. Journal of Financial education , 1-42. Bellovary, J. L., Giacomino, D. E., & Akers, M. D. (2007). A review of bankruptcy prediction studies: 1930 to present. Journal of Financial education , 1-42. Bentéjac, C., Csörgő, A., & Martínez-Muñoz, G. (2021). A comparative analysis of gradient boosting algorithms. Artificial Intelligence Review , 54 , 1937-1967. Biau, G. (2012). Analysis of a random forests model. The Journal of Machine Learning Research , 13 , 1063-1095. Bragoli, D., Ferretti, C., Ganugi, P., Marseguerra, G., Mezzogori, D., & Zammori, F. (2022). Machine-learning models for bankruptcy prediction: do industrial variables matter?. Spatial Economic Analysis , 17 (2), 156-177. Breiman, L. (2001). Random forests. Machine Learning, 45(1), 5–32. Chaudhuri, A., & De, K. (2011). Fuzzy support vector machine for bankruptcy prediction. Applied Soft Computing , 11 (2), 2472-2486. Chen, T., & Guestrin, C. (2016). Xgboost: A scalable tree boosting system. In Proceedings of the 22nd acm sigkdd international conference on knowledge discovery and data mining (pp. 785-794). Du Jardin, P. (2010). Predicting bankruptcy using neural networks and other classification methods: The influence of variable selection techniques on model accuracy. Neurocomputing, 73(10-12):2047–2060. Du Jardin, P. (2015). Bankruptcy prediction using terminal failure processes. European Journal of Operational Research , 242 (1), 286-303. Duénez-Guzmán, E. A., & Vose, M. D. (2013). No free lunch and benchmarks. Evolutionary Computation, 21 (2), 293-312. Edmister, R. O. (1972). An empirical test of financial ratio analysis for small business failure prediction. Journal of Financial and Quantitative analysis , 7 (2), 1477-1493. Hand, D., Mannila, H., & Smyth, P. (2001). Principles of data mining. Cambridge, MA: MIT Press. Huang, Z., Chen, H., Hsu, C. J., Chen, W. H., & Wu, S. (2004). Credit rating analysis with support vector machines and neural networks: a market comparative study. Decision support systems , 37 (4), 543-558. Imandoust, S. B., & Bolandraftar, M. (2013). Application of k-nearest neighbor (knn) approach for predicting economic events: Theoretical background. International journal of engineering research and applications , 3 (5), 605-610. Kasgari, Ahmad Ahmadpour, Seyyed Hasan Salehnezhad, and Fatemeh Ebadi. "The bankruptcy prediction by neural networks and logistic regression." International Journal of Academic Research in Accounting, Finance and Management Sciences 3.4 (2013): 146-152. Liang, D., Lu, C. C., Tsai, C. F., & Shih, G. A. (2016). Financial ratios and corporate governance indicators in bankruptcy prediction: A comprehensive study. European journal of operational research , 252 (2), 561-572. Mai, F., Tian, S., Lee, C., & Ma, L. (2019). Deep learning models for bankruptcy prediction using textual disclosures. European journal of operational research , 274 (2), 743-758. Min, J. H., & Lee, Y. C. (2005). Bankruptcy prediction using support vector machine with optimal choice of kernel function parameters. Expert systems with applications , 28 (4), 603-614. Maione, C., Batista, B.L., Campiglia, A.D., Barbosa, F., Jr, & Barbosa, R.M. (2016). Classification of geographic origin of rice by data mining and inductively coupled plasma mass spectrometry. Computers and Electronics in Agriculture , 121, 101–107. Muslim, M. A., & Dasril, Y. (2021). Company bankruptcy prediction framework based on the most influential features using XGBoost and stacking ensemble learning. International Journal of Electrical and Computer Engineering (IJECE) , 11 (6), 5549-5557. Narvekar, A., & Guha, D. (2021). Bankruptcy prediction using machine learning and an application to the case of the COVID-19 recession. Data Science in Finance and Economics , 1 (2), 180-195. Odom, M. D., & Sharda, R. (1990). A neural network model for bankruptcy prediction. In 1990 IJCNN International Joint Conference on neural networks (pp. 163-168). IEEE. Odom, M. D., & Sharda, R. (1990). A neural network model for bankruptcy prediction. In 1990 IJCNN International Joint Conference on neural networks (pp. 163-168). IEEE. Ohlson, J. A. (1980). Financial ratios and the probabilistic prediction of bankruptcy. Journal of accounting research , 109-131. Olson, D. L., Delen, D., & Meng, Y. (2012). Comparative analysis of data mining methods for bankruptcy prediction. Decision Support Systems , 52 (2), 464-473. Olson, D. L., Delen, D., and Meng, Y. (2012). Comparative analysis of data mining methods for bankruptcy prediction. Decision Support Systems, 52(2):464–473. Perboli, G., & Arabnezhad, E. (2021). A Machine Learning-based DSS for mid and long-term company crisis prediction. Expert Systems with Applications , 174 , 114758. Ramraj, S., Uzir, N., Sunil, R., & Banerjee, S. (2016). Experimenting XGBoost algorithm for prediction and classification of different datasets. International Journal of Control Theory and Applications , 9 (40), 651-662. Sandri, M., & Zuccolotto, P. (2008). A bias correction algorithm for the Gini variable importance measure in classifcation trees. Journal of Computational and Graphical Statistics, 17 (3), 611–628. Shrivastava, S., Jeyanthi, P. M., & Singh, S. (2020). Failure prediction of Indian Banks using SMOTE, Lasso regression, bagging and boosting. Cogent Economics & Finance, 8 (1), 1729569. Shetty, S., Musa, M., & Brédart, X. (2022). Bankruptcy Prediction Using Machine Learning Techniques. Journal of Risk and Financial Management , 15 (1), 35. Smith, M., & Alvarez, F. (2021). Predicting firm-level bankruptcy in the Spanish economy using extreme gradient boosting. Computational Economics , 1-33. Son, H., Hyun, C., Phan, D., & Hwang, H. J. (2019). Data analytic approach for bankruptcy prediction. Expert Systems with Applications , 138 , 112816. Tian, S. and Yu, Y. (2017). Financial ratios and bankruptcy predictions: An international evidence. International Review of Economics & Finance, 51:510–526. Tsai, C. F. (2008). Financial decision support using neural networks and support vector machines. Expert Systems , 25 (4), 380-393. Tsai, C. F., Hsu, Y. F., & Yen, D. C. (2014). A comparative study of classifier ensembles for bankruptcy prediction. Applied Soft Computing , 24 , 977-984. Tseng, F. M., & Hu, Y. C. (2010). Comparing four bankruptcy prediction models: Logit, quadratic interval logit, neural and fuzzy neural networks. Expert systems with applications , 37 (3), 1846-1853. Yeh, C.C., Chi, D.J., & Lin, Y.R. (2014). Going-concern prediction using hybrid random forests and rough set approach. Information Sciences, 254 , 98-110. Wang, G., Ma, J., & Yang, S. (2014). An improved boosting based on feature selection for corporate bankruptcy prediction. Expert Systems with Applications , 41 (5), 2353-2361. Wang, G., Ma, J., Huang, L., & Xu, K. (2012). Two credit scoring models based on dual strategy ensemble trees. Knowledge-Based Systems , 26 , 61-68. Wilson, R. L., & Sharda, R. (1994). Bankruptcy prediction using neural networks. Decision support systems , 11 (5), 545-557. Whitrow, C., Hand, D.J., Juszczak, P., Weston, D. & Adam, N.M. (2009). Transaction aggregation as a strategy for credit card fraud detection. Data Mining and Knowledge Discovery, 18 , 30-55. Xie, E., Li, X., Ngai, E. & Ying, W. (2009). Customer churn prediction using improved balanced random forest. Expert Systems with Applications, 36 , 5445-5449. Zhang, W. (2017). Machine learning approaches to predicting company bankruptcy. Journal of Financial Risk Management , 6 (04), 364. Zhao, H., Sinha, A. P., & Ge, W. (2009). Effects of feature construction on classification performance: An empirical study in bank failure prediction. Expert Systems with Applications , 36 (2), 2633–2644. Zoričák, M., Gnip, P., Drotár, P., & Gazda, V. (2020). Bankruptcy prediction for small-and medium-sized companies using severely imbalanced datasets. Economic Modelling , 84 , 165-176. Additional Declarations No competing interests reported. Supplementary Files Appendix.docx Cite Share Download PDF Status: Posted Version 1 posted 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-5693739","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":434070349,"identity":"8b4e10f7-2f82-4f8c-b0f2-ab7987a80cf4","order_by":0,"name":"Thu Hien Bui","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAvElEQVRIiWNgGAWjYBACA2YGBiCy4QfzeIjXkpAm2UO8FgawlsMkaDFnZz72uPDHeQl7iQTGB2/bGBK3E9Ji2cyWbjwj4bYEj0QCs+FcoJadDYQcdpjHTJon4XYdUAubNG8bg7HBAYJa+L8BtZwD2cL+m0gtPGxALQdAWtiYgVrkiNDCZiY9Iy1ZgufMw2bJOeckiNBy/vAz6QIbOwn29uSDH96U2fAQ1IIEGBuAhATx6kfBKBgFo2AU4AYAdwU1QEd4zHcAAAAASUVORK5CYII=","orcid":"","institution":"Foreign Trade University","correspondingAuthor":true,"prefix":"","firstName":"Thu","middleName":"Hien","lastName":"Bui","suffix":""},{"id":434070350,"identity":"1a4238c1-df96-407a-a5f9-2691121c2299","order_by":1,"name":"Thi Thuy Duong Truong","email":"","orcid":"","institution":"Banking Academy of Vietnam","correspondingAuthor":false,"prefix":"","firstName":"Thi","middleName":"Thuy Duong","lastName":"Truong","suffix":""},{"id":434070351,"identity":"d3fceb2b-1fad-46ff-86c5-43d11225081b","order_by":2,"name":"Thi Phuong Thao Tran","email":"","orcid":"","institution":"Foreign Trade University","correspondingAuthor":false,"prefix":"","firstName":"Thi","middleName":"Phuong Thao","lastName":"Tran","suffix":""}],"badges":[],"createdAt":"2024-12-22 12:38:09","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-5693739/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-5693739/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":79345715,"identity":"ef964b2c-d6ec-42ba-ad76-b99cc8a0d462","added_by":"auto","created_at":"2025-03-27 09:35:37","extension":"jpg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":184436,"visible":true,"origin":"","legend":"\u003cp\u003eCorrelation matrix of liquidity group’s variable\u003c/p\u003e\n\u003cp\u003e\u003cem\u003e\u003cstrong\u003eSource: By the authors\u003c/strong\u003e\u003c/em\u003e\u003c/p\u003e","description":"","filename":"1.jpg","url":"https://assets-eu.researchsquare.com/files/rs-5693739/v1/8bf564ee53c7559076767789.jpg"},{"id":79346798,"identity":"2be6fdac-cb4c-4316-9bdf-9ac1d5c97660","added_by":"auto","created_at":"2025-03-27 09:43:37","extension":"jpg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":40223,"visible":true,"origin":"","legend":"\u003cp\u003eThe important features by XGB\u003c/p\u003e\n\u003cp\u003e\u003cem\u003e\u003cstrong\u003eSource: By the authors\u003c/strong\u003e\u003c/em\u003e\u003c/p\u003e","description":"","filename":"2.jpg","url":"https://assets-eu.researchsquare.com/files/rs-5693739/v1/55392f40556bd3b0b7470a6a.jpg"},{"id":80227492,"identity":"0ddbb305-c4ea-4854-aafa-3cafa666afd3","added_by":"auto","created_at":"2025-04-09 11:54:07","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1458453,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-5693739/v1/0a2efa16-50a4-44ce-a4c5-2fbbd6fdac89.pdf"},{"id":79345718,"identity":"e824162f-f54a-4aae-b062-06a2de148ff2","added_by":"auto","created_at":"2025-03-27 09:35:37","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":233848,"visible":true,"origin":"","legend":"","description":"","filename":"Appendix.docx","url":"https://assets-eu.researchsquare.com/files/rs-5693739/v1/468334df89a2caf7ff107f60.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Financial ratios as indicators in bankruptcy prediction: A comparative analysis of statistical and machine learning models","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eBankruptcy prediction is critical to identifying potential risks of failure for the firms and other stakeholders such as investors, financial institutions, and governments (Zywicki, 2008; Liang et al., 2014). The capital markets have been growing sharply these days, and continually breaking previous peaks requires serious concentration on preventing and mitigating financial frauds and crises. Although the growth of the capital market presents many opportunities, it is also accompanied by risks for the firms and their stakeholders. Therefore, many researchers have employed different techniques to accurately estimate the likelihood of firm failure. The models are categorized into two types, namely statistical and machine learning techniques. Nevertheless, as a consequence of the market's growth in scale and complexity, predictions of statistical models are questioned. In the meantime, machine learning was utilized for better data processing and superior predictive model construction. Machine learning models are proven to have solved predictive errors recognized in traditional methods; hence, they are more favored in recent studies on bankruptcy prediction.\u003c/p\u003e \u003cp\u003eLiterature relies primarily on financial ratios computed from financial statements to forecast bankruptcy and financial distress. Besides, few studies add firms' credit attributes and characteristics as inputs for predictive models. However, financial ratios, which are complex information, are believed to be more objective, stable inputs for prediction. According to Liang et al. (2014), financial ratios can be classified into seven categories: solvency, profitability, cash flow ratios, capital structure ratios, turnover ratios, and growth. However, selecting these ratios for predictive models is still debated. According to Bellovary et al. (\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2007\u003c/span\u003e), there is a probability that a model with fewer ratios generates a more precise forecast compared to those that capture a lot. Few studies also demonstrate differences in the predictive power of distinct financial groups (Smith and Alvarez, \u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Son et al., \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). However, only a few findings generally cover the relevance of financial ratio groups in predicting bankruptcy. The development of intelligent methods and new approaches raises the performance of failure risk forecasting models (Du\u0026eacute;nez-Guzm\u0026aacute;n and Vose, \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2013\u003c/span\u003e). This is very important due to the selection of bankruptcy-predicting models depending on the firm's characteristics and data availability.\u003c/p\u003e \u003cp\u003eThe above gaps bring us to set up two separate challenges for predicting bankruptcy: The first task is to forecast the company's financial distress status using financial ratios two and one year before the event. In this task, we employ four statistical and machine learning models. Regarding statistical models, we use Linear Regression (LR) while Random Forest (RF), Extreme Gradient Boosting (XGB), and K-Nearest Neighbor (KNN) are used as machine learning models. The accuracy rate of each model in two periods will be computed to conclude which models are the most superior for the Vietnamese market and which other emerging markets share the same characteristics. The study identifies XGB (Extreme Gradient Boosting) as the most influential predictive model, followed by RF (Random Forest), and LR (Logistic Regression) as the least efficient. This highlights the effectiveness of machine learning techniques like XGB and RF in predicting bankruptcy in emerging markets, which might have been underexplored previously in the context of Vietnam. In addition, the study evaluates the accuracy of bankruptcy prediction models over different time horizons and finds that XGB and RF remain highly accurate up to two years ahead of the possibility of bankruptcy. Moreover, novel contributions of our study include the use of F1 Score, AUC-ROC, Brier Score, and Log-loss to enhance probability evaluation, alongside hyperparameter tuning, class imbalance handling, and grid search optimization to improve model performance. Finally, results confirm the relevance of all six financial indicator groups, with key ratios such as Total Assets Turnover, Current Assets to Sales, Current Assets to Total Assets, Equity to Total Assets and Total Equity Growth play a crucial role. These variables emphasize that a firm\u0026rsquo;s ability to manage its liquidity, efficiently utilize assets, and sustain capital growth plays a decisive role in determining its financial resilience. The integration of such indicators into advanced predictive models like XGBoost not only enhances predictive accuracy but also provides practical guidance for stakeholders, particularly financial institutions, investors, and corporate managers in developing early warning systems and strategic interventions to mitigate bankruptcy risk in the Vietnamese business context.\u003c/p\u003e \u003cp\u003eThis study is divided into six sections. This beginning part is an introduction, the first section, followed by the second section, where prior studies are reviewed to uncover research gaps and facilitate the construction of this study's research direction. Subsequently, the third section, Methodology, will provide details on research models. The fourth section clarifies the data used in the study. The fifth section presents findings before the general summary in the final quarter.\u003c/p\u003e"},{"header":"2. Literature review","content":"\u003cp\u003eBankruptcy refers to a firm's inability to accomplish its financial obligations, accompanied by a declaration of default by the court ruling. The event of bankruptcy risk or financial distress poses an excellent threat to the firm and its stakeholders. Although regulations and laws have been reinforced to ease the potential damages, the growing complexity of businesses and financial involvements, as well as the lack of a mature theoretical framework on corporate failure (Wang et al., \u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e2014\u003c/span\u003e), voice a need for exploration on determinant factors along with highly-accurate predictive models. In financial management, credit risk is minimized by assessing firms' economic well-being, which signals the likelihood of financial distress and bankruptcy. These assessments and forecasts act as crucial inputs for the decision-making process of firms' managers, investors, creditors, and policymakers (Zywicki, 2008; Liang et al., 2014).\u003c/p\u003e \u003cp\u003eAcknowledging the vital role of bankruptcy prediction, since the late 19th century, numerous researchers have conducted to develop and propose predictive models on bankruptcy to suggest early solutions and prevent detrimental effects. The literature mainly uses financial ratios computed from annual financial statements as critical indicators for bankruptcy (Mai et al., 2018). However, few studies also employ financial or credit attributes such as credit scores, company characteristics, etc., as inputs for their research models. Regarding research techniques, they are categorized into two types: traditional models and machine learning ones. Specifically, traditional models focus on statistical methods, among which the multivariate discriminant analysis model (MDA) and logistic regression model (LR) are the most popular and widely used methods in both academic studies and practice (Ohson, 1980; Ohson et al., 2012). Machine learning models are also termed artificial intelligence models in which data are processed and used for training and testing models, ensuring higher prediction validity compared to statistical methods.\u003c/p\u003e \u003cp\u003eThe early days of bankruptcy prediction officially began with the study of Beaver (\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e1966\u003c/span\u003e) using a linear regression model to classify failed and non-failed firms. The author employed 30 financial ratios specified into six categories: Cash flow ratios, Net income ratios, Debt to total asset ratios, Liquid asset to total asset ratios, Liquid asset to current debt ratios, and Turnover ratios. Noticeably, Beaver (\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e1966\u003c/span\u003e) examined the firms' performances five years before their failure, concluding that predictive power is reduced as time goes backward. Besides, the study also suggested that not all ratios affect bankruptcy similarly. Cash flow ratios are the most influential, while liquid asset ratios are the weakest determinants. In 1968, Altman proposed the best-known and most widely-used function to assess a firm's financial health. The study utilized a multiple discriminant analysis method (MDA) to build a process to compute a firm's Z-score based on five financial ratios in each profitability, activity, liquidity, solvency, and leverage ratio classification. The conclusions agreed with Beaver (\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e1966\u003c/span\u003e) regarding the predictive accuracy over time, which led Altman (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e1968\u003c/span\u003e) to recommend that prediction should be made in less or equal to 2 years. A study by Edmister (\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e1972\u003c/span\u003e) also applied MDA in exploring 19 financial ratios to predict the bankruptcy of small businesses. MDA is the most popular method utilized in the study of bankruptcy. In 1980, the logit analysis technique, particularly maximum likelihood estimation, was used by Ohlson (\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e1980\u003c/span\u003e) to uncover the effects of 6 ratios on the possibility of failure. The author also suggested further discoveries on other predictors.\u003c/p\u003e \u003cp\u003eNonetheless, since the 1990s, traditional techniques have been proven to be outperformed by modern models (Alaka, 2018; Du Jardin, \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2015\u003c/span\u003e; Kasgari et al., \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2013\u003c/span\u003e; Mai et al., \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Min and Lee, \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2005\u003c/span\u003e; Ohson et al., 2012; Ohson and Sharda, 1990; Wang et al., \u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e2012\u003c/span\u003e; Wilson and Sharda, \u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e1994\u003c/span\u003e), artificial intelligence (AI) have been introduced as an alternative for assessing firms' well-being, and predicting corporate failure. As one of the earliest studies on the use of machine learning to predict bankruptcy, Odom and Sharda (\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e1990\u003c/span\u003e) made a comparison on the accuracy of MDA and neural network technique (NN) in forecasting bankruptcy using the same five ratios as Altman's (1968). It was found that neural networks were more robust and had a higher accuracy rate than MDA, which lay a foundation for the application of modern techniques. Studies by Kasgari et al. (\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2013\u003c/span\u003e) and Son et al. (\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e2019\u003c/span\u003e) also investigated the performance of neural networks over the logistic regression method. The results demonstrated the superiority of machine learning methods over statistical ones. The rapid development of data mining presents many machine-learning models. Hence, studies have been conducted with several models to compare and decide on the most potent model. Besides, ensembles of various machine learning models such as Random Forest (RF) or Extreme Gradient Boosting (XGB) are also suggested (Perboli and Arabnezhad, \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Son et al., \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Wang et al., 2011) which are proven to strengthen predictive power. The table below demonstrates past studies using machine learning models with the best-performed models in bold:\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 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003ePast machine learning studies on bankruptcy prediction\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\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 \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNo.\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eWork\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eModels\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eRatio groups\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eDataset\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eOdom and Sharda (\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e1990\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eMDA, \u003cb\u003eNN\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eAltman\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eUS\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eWilson and Sharda (\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e1994\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003eSNN\u003c/b\u003e, MDA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eAltman\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eUS\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eHuang et al. (\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2004\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003eSVM\u003c/b\u003e, BPN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eProfitability, leverage, liquidity, actitivy, turnover ratio\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eUS, Taiwan\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMin and Lee (\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2005\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003eSVM\u003c/b\u003e, MDA, LR, SNN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eProfitability, leverage, liquidity, actitivy, turnover ratio\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eKorea\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTsai (\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2008\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eANN, \u003cb\u003eMLP\u003c/b\u003e, SVM\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eUS, Australia, Taiwan, Korea, German\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTseng and Hu (\u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2010\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eLR, QIL, MLP, \u003cb\u003eRBFN\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eManagement inefficiency, capital structure, insolvency, adverse economic effects, income volatility,\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eUK\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eChaudhuri and De (\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2011\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSVM, \u003cb\u003eFSVM\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eProfitability, leverage, liquidity, actitivy, turnover ratio\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eUS\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eOhson et al. (2012)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eANN, \u003cb\u003eDT\u003c/b\u003e, SVM, LR\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eAltman\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eUS\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eWang et al. (\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e2012\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003eRF\u003c/b\u003e, LR, MDA, MLP, RBFN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eAttributes not ratios\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eAustralia, German\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e10\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eKasgari et al. (\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2013\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003eANN\u003c/b\u003e, LR\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eLiquidity, profitability, leverage\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eIran\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e11\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTsai et al. (\u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e2014\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSVM, MLP, \u003cb\u003eDT\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eCredit data sets\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eAustralia, German, Japan\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e12\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eDu Jardin (\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2015\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eMDA, LR, \u003cb\u003eMLP\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eLiquidity, solvency, profitability, financial structure, activity, turnover\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eFrance\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e13\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eLiang et al. (\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2016\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003eSVM\u003c/b\u003e, KNN, CART, MLP, NB\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eSolvency, capital structure, profitability, turnover ratios, cash flow ratios, growth, corporate governance indicators\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eTaiwan\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e14\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMai et al. (\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2019\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eLR, SVM, \u003cb\u003eRF\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eLiquidity, profitability, leverage\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eUS\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e15\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSon et al. (\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e2019\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003eRF\u003c/b\u003e, DT, XGB, KNN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eCash flow ratios, growth, leverage, liquidity, activity, profitability, solvency\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eKorea\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e16\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMuslim and Dasril (\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2021\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eKNN, DT, SVM, \u003cb\u003eRF, XGB\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003ePoland\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e17\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNarvekar and Guha (\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2021\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eRF, SVM, \u003cb\u003eXGB\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eUS\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e18\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePerboli and Arabnezhad (\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2021\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eRF, \u003cb\u003eXGB\u003c/b\u003e, LR, NN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eProfitability, actitivy, leverage, liquidity\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eItaly\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e19\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSmith and Alvarez (\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e2021\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eLR, SNN, \u003cb\u003eRF\u003c/b\u003e, SVM, XGB,\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eProfitability, actitivy, leverage, liquidity, turnover\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eSpain\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e20\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eShetty et al. (\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e2022\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNN, \u003cb\u003eSVM\u003c/b\u003e, XGB\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eLiquidity, solvency, turnover\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eBelgium\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e21\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eBragoli et al. (\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2022\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eLR, NN, RF, \u003cb\u003eXGB\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eProfitability, liquidity, growth\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eItaly\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 \u003cem\u003e* ANN: Artificial Neutral Networks; BPN: Back-Propagation Neural Networks; CART: Classification And Regression Tree; CBR: Case-Based Reasoning; DT: Decision Trees; IF: Isolation Forest; KNN: K-Nearest Neighbor; LDA: Linear Discriminant Analysis; LR: Logistic Regression; LSAD: Least-Squares Anomaly Detection; MLP: Multi-Layer Perceptron; NB: Naive Bayes; NN: Neural Networks; OCSVM: One Class SVM; QIL: Quadratic Interval Logit Model; RBFN: Radial Basis Function Network; SAT: Single Attribute Test; SNN: Shallow Neutral Network; SVM: Support Vector Machines; XGB: Extreme Gradient Boosting.\u003c/em\u003e \u003c/p\u003e \u003cp\u003e \u003cb\u003eSource: By the authors\u003c/b\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 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003ePredictive accuracy by model\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\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=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNo.\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eModel\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eLowest accuracy\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eHighest accuracy\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eStudy with highest accuracy\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eRF\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e63.90%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e99.68%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eNarvekar and Guha (\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2021\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eXGB\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e81.00%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e98.70%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eNarvekar and Guha (\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2021\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eKNN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e60.20%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e95.60%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eMuslim and Dasril (\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2021\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eDT\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e73.33%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e94.80%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eMuslim and Dasril (\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2021\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSVM\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e66.10%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e90.10%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eNarvekar and Guha (\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2021\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMLP\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e61.40%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e93.75%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eTseng and Hu (\u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2010\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e71.00%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e98.00%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eWilson and Sharda (\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e1994\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eLR\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e49.10%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e90.80%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eHuang et al. (\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2004\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMDA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e37.50%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e95.60%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eWilson and Sharda (\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e1994\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"5\"\u003e\u003cb\u003eSource: By the authors\u003c/b\u003e\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eIn general, it is noticed that the later the studies are, the more popular the ensemble techniques are, of which superiority is evidenced (Mai et al., \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Perboli and Arabnezhad, \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). Studies on bankruptcy prediction frequently mention Type I and Type II errors. These are errors regarding the misclassification of the firms\u0026rsquo; financial situation. Type I errors wrongly forecast bankrupt firms as non-bankrupt, while Type II errors happen when non-bankrupt firms are misclassified as bankrupt. Type I is argued to be the more costly of the two types of error because it poses more damages regarding a firm\u0026rsquo;s reputation, loss of its shareholders, and potential court costs (Bellovary et al., \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2007\u003c/span\u003e). Hence, although studies try to handle both types of errors, models reducing the frequency of Type I errors are more favoured.\u003c/p\u003e \u003cp\u003eAmong the models used, RF, XGB, and KNN are the most popular and accurate methods used to construct predictive models on firm failure. These models are known to have resolved the issues of data dimensionality, scalability, and accuracy of classifications which confronted by previous models (Belgiu and Dragut, 2016; Bentejac et al., 2021; Biau, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2012\u003c/span\u003e; Breiman, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2001\u003c/span\u003e; Chen and Guestrin, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2016\u003c/span\u003e; Ramraj et al., \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2016\u003c/span\u003e; Zhang, \u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). Regarding the data captured by the models, the majority of researchers use financial ratios classified as listed in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e. In short, ratios uncovering firms\u0026rsquo; liquidity, profitability, cash flow, capital structure, turnover, and growth are all studied. Nevertheless, only several studies comment on the differences in the predictive power of specific financial categories. A study by Son et al. (\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e2019\u003c/span\u003e) suggests that among six groups, liquidity, solvency, and growth ratios are more relevant to bankruptcy forecast, while leverage and capital structure ratios are proven to be closely related to predictive performance, according to Smith and Alvarez (\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). In contrast, Liang et al. (\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2016\u003c/span\u003e) conclude the crucial roles of solvency and profitability ratios in the prediction. In addition, machine learning models provide quite high results for the bankruptcy prediction problem (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e, see for example).\u003c/p\u003e \u003cp\u003eThis study extends the bankruptcy prediction literature in three ways. Firstly, the most popular and accurate machine learning models, inlcuding RF, XGB, and KNN will be utilized in this study. Moreover, a statistical model, namely LR, will also be run to compare and decide which model is the most suitable for the Vietnamese market and other similar emerging markets. Secondly, we also input 62 financial ratios classified into six groups to strengthen the predictive power of the models. Lastly, the influences of ratio groups on forecasting firm failure are examined to determine which groups are more correlated with bankruptcy prediction.\u003c/p\u003e"},{"header":"3. Methodology","content":"\u003cp\u003eThis study uses supervised machine learning models, including Random Forest (RF), Extreme Gradient Boosting (XGB), and K-Nearest Neighbor (KNN), to construct a function to make predictions on the health of firms.\u003c/p\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003e3.1 Random Forest\u003c/h2\u003e \u003cp\u003eRandom Forest is an ensemble bagging tree-based learning algorithm introduced by Breiman (\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2001\u003c/span\u003e). The mechanism of the ensemble model is to integrate different classifiers to create novel ones more efficiently and accurately. The Random Forest Classifier is constructed using decision trees from bootstrapped data, trees have been trained using randomly chosen subsets of both the training set and the feature set.\u003c/p\u003e \u003cp\u003eThe following steps perform the random forest: Let the data be n observations and p features.\u003c/p\u003e \u003cp\u003eStep 1. From the train data, randomly bootstrapped sample set of size n.\u003c/p\u003e \u003cp\u003eStep 2. Select randomly k variables from p features.\u003c/p\u003e \u003cp\u003eStep 3. Create the decision trees from the collected data.\u003c/p\u003e \u003cp\u003eIn light of the k feature vector, each tree in the collection casts a vote to categorize the sample. The final outcome obtained the majority of votes.\u003c/p\u003e \u003cp\u003eThe Random Forest approach uses randomly selected training data and attribute subsets; each decision tree may be weak in classification and result in a significant degree of bias. However, the method's end product is an aggregate of multiple decision trees, the information from the trees will complement one another and provide a model with accurate prediction outcomes. The ability to deal with outliers and noise is the advantage of RF (Yeh et al., \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2014\u003c/span\u003e). Besides classification and prediction, it can determine the importance of variables in the model, which enhances the model's performance and reduces the data dimension (Maione et al., \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2016\u003c/span\u003e). In the financial industry, Random forest has been used to detect credit fraud (Whitrow et al., \u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e2009\u003c/span\u003e) and predict customer churn using bank services (Xie et al., \u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e2009\u003c/span\u003e), and bankruptcy prediction (Zorič\u0026aacute;k et al., \u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e2020\u003c/span\u003e).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003e3.2 Extreme Gradient Boosting\u003c/h2\u003e \u003cp\u003eExtreme Gradient Boosting is an ensemble method called boosting, which has been widely used recently because of its dominance (Barboza et al., \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). Models are continuously trained to allow each model to get better and make up for the shortcomings of the preceding one. Different boosting methods develop and aggregate weak learners differently during the sequential stacking process. XGB, proposed by Fiedman (2001), is a distributed gradient-boosting library developed to be very effective, adaptable, and portable. By integrating a convex loss function based on the difference between the expected and actual outputs and a penalty term for model complexity, XGB minimizes a regularized (L1 and L2) objective function. To create the final forecast, past trees are blended with residuals or errors from earlier trees as the training process continues.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec6\" class=\"Section2\"\u003e \u003ch2\u003e3.3 K-Nearest Neighbor\u003c/h2\u003e \u003cp\u003eKNN is one of the most straightforward classifier algorithms depending on the distance (Zhao et al., \u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e2009\u003c/span\u003e). The distance measures the similarity between the classified set and the training set. In the KNN algorithm, after determining the number of k nearest neighbors, the algorithm finds out the k-labeled sets that are nearest the unlabeled partner, and the partner is assigned a label by majority vote amongst its k-nearest neighbors (Hand et al., \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2001\u003c/span\u003e; Zhao et al., \u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e2009\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThe KNN algorithm is straightforward to utilize as it only depends on the popular distances. The technique is implemented without making any assumptions about how data are distributed. The methodology is memory-based, and new training data is updated immediately after each classification step. The number of k-neighbors is delicate, affecting the testing model's performance.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec7\" class=\"Section2\"\u003e \u003ch2\u003e3.4 Logistic regression\u003c/h2\u003e \u003cp\u003eThe logistic regression model is one of the traditional statistical methods that investigates the relationship between explanatory variables and dependent variables. Many studies were effectuated to forecast the health of companies (Du Jardin, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2010\u003c/span\u003e; Olson et al., \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2012\u003c/span\u003e; Maalouf and Trafalis, 2011; Kim, 2011; Tsai, 2014). The dependent variable in a logistic regression model may be binary or categorical, and the independent variables may be a combination of continuous, flat, and binary variables. The model is formed as\u003cdiv id=\"Equa\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equa\" name=\"EquationSource\"\u003e\n$$\\:z={\\beta\\:}_{0}+{\\beta\\:}_{1}{X}_{1}+{\\beta\\:}_{2}{X}_{2}+...+{\\beta\\:}_{m}{X}_{m},\\:y=\\frac{{e}^{z}}{1+{e}^{z}}$$\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003ewhere \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{X}_{1},...,{X}_{m}\\)\u003c/span\u003e\u003c/span\u003e are explanatory variables, the output belongs to (0,1), and the object is classified into a positive class if the output is greater than 0.5, else a negative class.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003e3.5 Performance comparison measures\u003c/h2\u003e \u003cp\u003eConfusion matrix\u003c/p\u003e \u003cp\u003eThe prediction task can be conceptualized as a binary prediction, where the bankruptcy class is represented as a positive status (1) and the health class as a negative status (0). We use various measures to evaluate the effectiveness of the forecasting model. The confusion matrix is displayed in Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e (Son et al., \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Shrivastava et al., \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2020\u003c/span\u003e).\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 3\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eConfusion matrix\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"4\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c4\" namest=\"c3\"\u003e \u003cp\u003eActual class\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003ePredicted class\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eTrue positive (TP)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eFalse positive (FP)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eFalse negative (FN)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eTrue negative (TN)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"4\"\u003e\u003cb\u003eSource: By the authors\u003c/b\u003e\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eThe positive class is anticipated to be positive, according to the true positive (TP). The same idea is known as true negative for actual negative class (TN). False positives (FP) are data points forecasted as positive in negative actual class. When the actual value is positive to be negative forecasted, the outcome is categorized as false negative (FN). The model's performance is determined via accuracy, precision and recall criteria. They stem from the confusion matrix, which is described in Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e.\u003cdiv id=\"Equb\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equb\" name=\"EquationSource\"\u003e\n$$\\:\\text{Accuracy}=\\frac{TP+TN}{TP+FP+TN+FN}$$\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003eThe proper forecast ratio for positive and negative classes is extracted by accuracy, a crucial performance metric. However, the number of firms classified into two categories is imbalanced, and the percentage of correct prediction needs to be increased to evaluate the efficient model. We also use additional helpful measures that, depending on the target stakeholders, can be used to describe the performance of the models. These are the precision, and recall for each class, which are defined as follows:\u003cdiv id=\"Equc\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equc\" name=\"EquationSource\"\u003e\n$$\\:\\text{Precision}=\\frac{TP}{TP+FP},\\:\\text{Recall}=\\frac{TP}{TP+FN}$$\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003ePrecision is the percentage of data items that are genuinely positive compared to those that the model has classified as positive (TP\u0026thinsp;+\u0026thinsp;FP). Recall displays the proportion of actual positive data points to those that are positive (TP\u0026thinsp;+\u0026thinsp;FN).\u003c/p\u003e \u003cp\u003eIn addition, to evaluate the models, we also use F1 score, AUC-ROC, log- loss and brier score. The F1 score shows the model\u0026rsquo;s performance, it balances precision and recall in classification models. AUC-ROC binary classifier evaluates the discrimination between positive and negative classes. The log-loss is used to assess a classification model's probability-based predictions. It gauges how closely the actual class labels match the predicted probability. The Brier score measures the accuracy of the predicted probability.\u003c/p\u003e \u003cp\u003e \u003cspan class=\"InlineEquation\"\u003e \u003c/span\u003e \u003c/p\u003e \u003cp\u003e \u003cspan class=\"InlineEquation\"\u003e \u003c/span\u003e \u003c/p\u003e \n\u003cp\u003e\u003cimg src=\"data:image/png;base64,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\" style=\"width: 625px; height: 178.957px;\" width=\"625\" height=\"178.957\"\u003e\u003c/p\u003e\u003cp\u003eA good model has excellent accuracy, precision, recall, F1 score, AUC-ROC and low log-loss, brier score.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec9\" class=\"Section2\"\u003e \u003ch2\u003e3.6 Dataset\u003c/h2\u003e \u003cp\u003eTo forecast the bankruptcy of firms in Vietnam, the data is selected from the primary financial sources available on Fiingroup.com from 2010 to 2021. The data is related to six financial indicator groups representing liquidity, capital structure, profitability, turnover ratio, cash flow ratios, and growth. We extracted and computed 62 financial criteria that presented as explanatory variables and the status of firms as dependent variables (failure and non-bankruptcy status firms, encoded the value one and zero, respectively). The definition of variables is described in Appendix 1 for more detail. After cleaning, processing and filtering data with missing values, the database is divided into two sets for two main tasks: one-year ahead bankruptcy prediction and two-year ahead bankruptcy prediction. The first task extracts financial variables of firms from 2010 to 2020 for the prediction of company status in the next year using the last fiscal year available. For example, for 2010, the machine learning models train and fit their models using explanatory variables of that year for prediction in 2011. The second set includes variables from 2010 to 2019 to forecast the health of a company in the next two years. For example, the descriptive statistics of the first set\u0026rsquo;s independent variables are listed in Appendix 2\u0026ndash;7. As summarized in Appendix 2\u0026ndash;7, most financial ratios witness a significant spread of their data. In other words, the firms studied in this work have quite different characteristics.\u003c/p\u003e \u003cp\u003eThe ensemble approach does not set strict conditions as conventional regressions do. If the variables in the model have a strong correlation, traditional regression can result in the multicollinearity problem. As a result, the prediction's accuracy is decreased. Figure\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e shows the correlation of variables related to firm liquidity. There are some high correlation variables such as the current ratio and quick ratio, current ratio and quick assets/current liability, total liabilities and total assets, and current liability to total assets. However, in intelligent methods like RF, XGB can handle the multicollinearity using decision trees. The models only use a few regressors to enhance prediction accuracy and resist multicollinearity issues (Sandri and Zuccolotto, \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2008\u003c/span\u003e). Appendix 8\u0026ndash;12 show the high correlation of variables, to name a few. In addition, there is a high correlation between all explanatory variables without affecting the effectiveness of the models.\u003c/p\u003e "},{"header":"4. Results and Discussions","content":"\u003cp\u003eIn order to accomplish two tasks for one-year ahead and two-year ahead prediction, each dataset is split into two data sets for the training and testing model using cross-validation methods. We repeat the number of fold for the best performance, the suitable number of k folds are 3 and 4 for one-year ahead and two-year ahead.\u003c/p\u003e \u003cp\u003eThe training sets are used to train the model, including Random Forest (RF), K-Nearest Neighbor (KNN), Extreme Gradient Boosting (XGB), and Logistic Regression (LR). Training outcomes will be exerted to test performance and prediction. The first training data set is used to train the forecasting model healthy status of firms for the next year using the previous year\u0026rsquo;s data, the health of firms in the next two years is predicted based on the second training data.\u003c/p\u003e \u003cp\u003eAll models use the same training set for binary classification concerning positive and negative classes. To optimize the performance of the models, we set hyperparameters to find learning rate, the depth of trees, minimal samples in leaf and split, the number of features for splitting, class weight to avoid imbalance.\u003c/p\u003e \u003cp\u003eThe matrixes are represented in Table\u0026nbsp;\u003cspan refid=\"Tab5\" class=\"InternalRef\"\u003e5\u003c/span\u003e and Table\u0026nbsp;\u003cspan refid=\"Tab6\" class=\"InternalRef\"\u003e6\u003c/span\u003e for two tasks on the test sets.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab4\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 4\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003ePerformance of one-year ahead forecasting\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"8\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eModel\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAccuracy\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePrecision\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eRecall\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eF1-score\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eAUC-ROC\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003eLog-loss\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c8\"\u003e \u003cp\u003eBrier Score\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eKNN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.8834\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.8918\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.98\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.9338\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.8819\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.4059\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.0867\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eRF\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003e0.9428\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003e0.9591\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e0.9733\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003e0.9662\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e\u003cb\u003e0.9773\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e\u003cb\u003e0.149\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e\u003cb\u003e0.042\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eXGB\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.9369\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.9494\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.9757\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.9624\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.9757\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.1585\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.0463\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLR\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.8465\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.972\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.841\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.902\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.9335\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.3901\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.1075\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"8\"\u003e\u003cb\u003eSource: By the authors\u003c/b\u003e\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab5\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 5\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003ePerformance of two-year ahead forecasting\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"8\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eModel\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAccuracy\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePrecision\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eRecall\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eF1-score\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eAUC-ROC\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003eLog-loss\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c8\"\u003e \u003cp\u003eBrier Score\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eKNN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.869\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.873\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.983\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.925\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.872\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.46\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.096\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRF\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.931\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.945\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.975\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.959\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.963\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.184\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.053\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eXGB\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003e0.933\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003e0.953\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e0.967\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003e0.96\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e\u003cb\u003e0.966\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e\u003cb\u003e0.174\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e\u003cb\u003e0.05\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLR\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.905\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.933\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.953\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.942\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.936\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.274\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.071\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"8\"\u003e\u003cb\u003eSource: By the authors\u003c/b\u003e\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eAs shown in Tables\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e and \u003cspan refid=\"Tab5\" class=\"InternalRef\"\u003e5\u003c/span\u003e, the models accurately classify the firms into two classes. Confirming findings of prior studies (Bragoli et al., \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Muslim and Dasril, \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Narvekar and Guha, \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Perboli and Arabnezhad, \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2021\u003c/span\u003e), RF outperforms in comparison to RF, KNN, and XGB with accuracy, precision and recall, AUC-ROC, F1 socre of over 93 percent in one year ahead case, followed by XGB, KNN, while LR gets the lowest prediction. XGB is more efficient than RF for a two-year ahead set, however, this diference is not significant. So, RF and XGB outperformance in risk prediction.\u003c/p\u003e \u003cp\u003eThe outcomes show that it slightly decreases the predicted performance compared to two periods; the forecasting after one year is higher than two years ahead. In other words, the predictive performance is better as it comes nearer to the event of bankruptcy, which is similar to previous works (Mai et al., 2018; Zoricak et al., 2020; Narvekar and Guha, \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Smith and Alvarez, \u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e2021\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eWhile high Recall indicators illustrate a low probability of bankrupt firms being omitted, Precision measures the accuracy of prediction; in other words, the higher Precision is, the more accurate the forecast is, high AUC-ROC shows the significant classification, low log-loss and brier score indicate that the predicted probabilities are close to the actual outcomes. In short, a good model is one in which both Recall, Precision, AUC-ROC, F1 score are high. The results of the firm's risk prediction demonstrate the superior performance of the models, with high accuracy and strong discriminatory power. In an overall comparison, KNN produced the least favorable prediction results, with higher log-loss and Brier score compared to the other three models. Moreover, in all models, there are no significant changes regarding the accuracy rate, which demonstrates that bankruptcy can be precisely predicted up to two years before the event of default.\u003c/p\u003e \u003cp\u003eTo uncover the most correlated financial ratio groups with bankruptcy prediction, one-year ahead forecasts are run except for one particular group among liquidity ratios, capital structure ratios, profitability ratios, turnover ratios, cash flows ratios, and growth ratios. The results are demonstrated in Table\u0026nbsp;\u003cspan refid=\"Tab6\" class=\"InternalRef\"\u003e6\u003c/span\u003e below:\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab6\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 6\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003ePerformance of one-year ahead forecasting without a group of financial ratios\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"8\"\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 \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eModel\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAccuracy\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePrecision\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eRecall\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eF1-score\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eAUC-ROC\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003eLog-loss\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c8\"\u003e \u003cp\u003eBrier Score\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"8\" nameend=\"c8\" namest=\"c1\"\u003e \u003cp\u003e\u003cem\u003eWithout liquidity group\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eKNN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.879\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.888\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.98\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.932\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.879\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.423\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.088\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRF\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.941\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.959\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.971\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.965\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.975\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.151\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.043\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eXGB\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.944\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.962\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.971\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.967\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.979\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.145\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.041\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLR\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.924\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.949\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.962\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.955\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.945\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.259\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.06\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"8\" nameend=\"c8\" namest=\"c1\"\u003e \u003cp\u003e\u003cem\u003eWithout capital structure group\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eKNN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.898\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.909\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.976\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.942\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.909\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.336\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.078\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRF\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.9428\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.959\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.973\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.966\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.977\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.149\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.042\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eXGB\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.944\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.962\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.971\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.967\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.978\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.145\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.041\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLR\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.943\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.959\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.974\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.966\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.973\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.19\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.04\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"8\" nameend=\"c8\" namest=\"c1\"\u003e \u003cp\u003e\u003cem\u003eWithout profitability group\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eKNN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.887\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.894\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.981\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.936\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.896\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.339\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.083\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRF\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.946\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.967\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.969\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.968\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.98\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.132\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.039\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eXGB\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.943\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.96\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.973\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.966\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.979\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.159\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.042\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLR\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.9428\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.961\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.971\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.966\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.971\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.21\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.042\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"8\" nameend=\"c8\" namest=\"c1\"\u003e \u003cp\u003e\u003cem\u003eWithout turnover ratio group\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eKNN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.857\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.869\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.976\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.92\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.811\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.529\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.109\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRF\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.947\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.968\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.969\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.969\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.978\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.142\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.04\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eXGB\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.947\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.966\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.971\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.969\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.980\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.133\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.039\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLR\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.942\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.955\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.976\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.966\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.973\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.186\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.041\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"8\" nameend=\"c8\" namest=\"c1\"\u003e \u003cp\u003e\u003cem\u003eWithout cash flow ratio group\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eKNN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.892\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.98\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.938\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.897\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.349\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.082\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRF\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.947\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.964\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.973\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.969\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.979\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.135\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.039\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eXGB\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.944\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e9.962\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.971\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.967\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.978\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.145\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.04\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLR\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.942\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.962\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.962\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.966\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.971\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.215\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.042\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"8\" nameend=\"c8\" namest=\"c1\"\u003e \u003cp\u003e\u003cem\u003eWithout growth group\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eKNN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.886\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.893\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.981\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.935\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.885\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.374\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.086\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRF\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.941\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.96\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.969\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.965\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.976\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.149\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.042\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eXGB\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.943\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.959\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.974\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.966\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.979\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.158\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.042\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLR\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.926\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.953\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.959\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.956\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.953\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.239\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.055\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"8\"\u003e\u003cb\u003eSource: By the authors\u003c/b\u003e\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab6\" class=\"InternalRef\"\u003e6\u003c/span\u003e shows the risk prediction results when excluding a group of financial indicators. This implies that risk prediction should consider all six aspects of financial indicators. The results also demonstrate that the predictive power of financial groups may vary across different economies due to their nature and characteristics.\u003c/p\u003e \u003cp\u003eIn addition, the findings show that the most suitable model to effectively predict bankruptcy risk for Vietnamese businesses is XGBoost. These results indicate that the machine learning model has an advantage in forecasting problems compared to the traditional logit model. At the same time, \"boosting\" algorithms like XGBoost often outperform other intelligent methods for forecasting problems. The results align with prior research such as Narvekar and Guha (\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2021\u003c/span\u003e), Perboli and Arabnezhad (\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2021\u003c/span\u003e), Bragoli et al. (\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). To explore which features have the influence of financial distress prediction of firms, the study uses XGB musesod to calculate Fscore. It reflects the total number of times that variable appears during the creation of data division nodes across the entire ensemble of trees of the XGBoost model. The Fscore of variables is shown in Appendix 13 and Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e.\u003c/p\u003e\u003cp\u003eLooking at the Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e and Appendix 13, X40, X46, X6, X18, X60 which representative for Total Assets Turnover, Current Assets to Sales, Current Assets to Total Assets, Equity to Total Assets and Total Equity Growth have the most important roles in predicting the failure risk. The emphasis on these ratios highlights several key aspects. Firstly, Total Assets Turnover (X40) measures the efficiency of a company's use of its assets to generate sales. A higher ratio indicates better performance and efficiency, which is crucial for assessing the operational efficiency and effectiveness of asset utilization in predicting company success or failure. Secondly, Current Assets to Sales (X46) helps in understanding how well a company can support its sales with its current assets, which is important for short-term financial stability and liquidity. This is critical for assessing overall profitability and shareholder value. In addition, Current Assets to Total Assets (X6) shows the proportion of current assets in relation to total assets. It indicates how much of the firm's assets are in liquid form, which is essential for understanding the liquidity position and the ability to cover short-term obligations. Equity to Total Assets (X18) reflects the proportion of a company's assets financed by shareholders' equity rather than debt. This not only enhances financial stability but also lowers the risk of insolvency during periods of financial stress. Finally, Total Equity Growth (X60) reflexes financial heath of firms. The company has a strong ability to self-finance its operations, which reduces its dependence on external debt Therefore, it reduces financial risk. To wrap up, the variables X40, X46, X18, X6, and X60 collectively underline the importance of liquidity, turnover ratios, and growth in predicting the failure risk of Vietnamese firms. The reason due to the feature of financial system in Vietnam, which is the bank-based financial system. Therefore, companies that manage their turnover, maintain healthy profitability, and ensure sufficient liquidity are better positioned to avoid financial difficulties and sustain their operations in the long run.\u003c/p\u003e"},{"header":"5. Conclusions and limitations","content":"\u003cp\u003eTo forecast firm bankruptcies in Vietnam, data spanning from 2010 to 2021 was collected from primary financial sources available on Fiingroup.com. The dataset comprises six key financial indicator groups: liquidity, capital structure, profitability, turnover ratio, cash flow ratios, and growth. A total of 62 financial criteria were derived and calculated as explanatory variables, with the firms' statuses categorized as dependent variables (coded as one for failure and zero for non-bankruptcy status firms). For modeling purposes, we employed a combination of statistical and machine learning techniques. Linear Regression (LR) was utilized as the statistical model, while Random Forest (RF), Extreme Gradient Boosting (XGB), and K-Nearest Neighbor (KNN) were employed as machine learning models. The results have contributed crucially to the literature on bankruptcy prediction in an emerging market such as Vietnam. Specifically, the most suitable predictive model is chosen while the impacts of financial ratio groups on the forecast are examined and concluded. Regarding the models for predicting bankruptcy or financial distress, in other words, XGB is the most influential model, with an accurate rate of 93% in both periods. Ranking right behind XGB is RF, with over 93% accuracy, but is slightly lower. KNN is the least efficient model, which forecast 88.34% of bankrupt cases and the ability classification is 88.19%. Another finding to withdraw is that the accuracy slightly increases as it is nearer to the actual event. However, the accuracy rates are still high at 93%, which ensures strong predictability of XGB and RF up to two years ahead of the possibility of bankruptcy. This underscores the effectiveness of advanced machine learning techniques, which might have been underexplored previously.\u003c/p\u003e \u003cp\u003eUnlike previous studies that primarily rely on accuracy, precision, and recall, this research introduces additional evaluation metrics, including F1 score, AUC-ROC to assess performance and discriminative power, and Brier Score and Log-loss to evaluate the quality of probability predictions. Furthermore, the study employs model fine-tuning through hyperparameter optimization, addresses class imbalance by adjusting class weights, and applies grid search to identify the most effective model structure. These enhancements significantly improve predictive performance. In addition, empirical results confirm the importance of all six financial indicator groups in forecasting bankruptcy, while also highlighting that no single group, when omitted, leads to a significant drop in accuracy, indicating model stability. Among the 62 financial ratios analyzed, variables such as Total Assets Turnover, Current Assets to Sales, Current Assets to Total Assets, Equity to Total Assets and Total Equity Growth emerge as the most influential in predicting financial failure.\u003c/p\u003e \u003cp\u003eThe findings hold practical value for multiple stakeholders. For example, financial institutions can better assess credit risk and allocate resources more effectively, investors gain a more accurate tool for evaluating firm sustainability and managing portfolio risk, and corporate managers can use these insights to strengthen internal financial structures and prevent insolvency. By enhancing the reliability and interpretability of bankruptcy prediction models, this study supports more informed decision-making and risk management strategies, particularly within the dynamic and evolving landscape of the Vietnamese business environment.\u003c/p\u003e \u003cp\u003eThis study acknowledges some limitations. Firstly, although the study successfully uncovers that each financial ratio category can have different impacts on the forecast of bankruptcy, models still need to be adjusted to strengthen the accuracy rate by re-selecting ratios. Secondly, in addition to financial ratios, credit attributes and firm characteristics are also evidenced to affect predictive power; however, they are not studied in this study. Therefore, several future research directions are suggested. Despite the increasing accuracy of bankruptcy predictive models, practical application demands a more optimized set of inputs. As revealed by the findings, it is noticed that the rate increases accurately, nonetheless insignificantly, in the absence of certain ratio groups. In addition, credit attributes and firm characteristics may also be determinants of prediction. Finally, incorporating time-series models (e.g., LSTM, ARIMA, Hidden Markov Models) to account for the evolving nature of financial health is certainly an interesting direction. These suggestions pave the way for further advancements in bankruptcy prediction research, facilitating more informed decision-making within the Vietnamese and broader emerging market contexts.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e \u003cb\u003e*Ethical Approval and Consent to Participate\u003c/b\u003e \u003c/p\u003e \u003cp\u003eThe authors declare to follow the ethical guidelines and consent to participate.\u003c/p\u003e \u003cp\u003e \u003cb\u003e*Consent for Publication\u003c/b\u003e \u003c/p\u003e \u003cp\u003eThe authors give consent for the publication.\u003c/p\u003e \u003cp\u003e \u003cb\u003e*Funding\u003c/b\u003e \u003c/p\u003e \u003cp\u003eNot applicable\u003c/p\u003e\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eAuthor Contributions StatementConceptualization, B.T.H; methodology, T.T.T.D; validation, T.T.P.T ; writing\u0026mdash;original draft preparation, B.T.H; writing\u0026mdash;review and editing, T.T.T.D; visualization, T.T.P.T.AcknowledgementNot ApplicableFundingNot Applicable\u003c/p\u003e\u003ch2\u003eAcknowledgement\u003c/h2\u003e\u003cp\u003enot applicable\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eThe data that support the findings of this study are openly available at https://doi.org/10.7910/DVN/6B91QV, Harvard Dataverse, V1.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eAlaka, H. A., Oyedele, L. O., Owolabi, H. A., Kumar, V., Ajayi, S. O., Akinade, O. O., \u0026amp; Bilal, M. (2018). Systematic review of bankruptcy prediction models: Towards a framework for tool selection. \u003cem\u003eExpert Systems with Applications\u003c/em\u003e, \u003cem\u003e94\u003c/em\u003e, 164-184.\u003c/li\u003e\n\u003cli\u003eAltman, E. I. (1968). Financial ratios, discriminant analysis and the prediction of corporate bankruptcy. \u003cem\u003eThe journal of finance\u003c/em\u003e, \u003cem\u003e23\u003c/em\u003e(4), 589-609.\u003c/li\u003e\n\u003cli\u003eBarboza, F., Kimura, H. \u0026amp; Alman, E. (2017). Machine learning models and bankruptcy prediction. Expert Systems with Applications, 83, 405-417.\u003c/li\u003e\n\u003cli\u003eBeaver, W. H. (1966). Financial ratios as predictors of failure. \u003cem\u003eJournal of accounting research\u003c/em\u003e, 71-111.\u003c/li\u003e\n\u003cli\u003eBelgiu, M., \u0026amp; Drăguţ, L. (2016). Random forest in remote sensing: A review of applications and future directions. \u003cem\u003eISPRS journal of photogrammetry and remote sensing\u003c/em\u003e, \u003cem\u003e114\u003c/em\u003e, 24-31.\u003c/li\u003e\n\u003cli\u003eBellovary, J. L., Giacomino, D. E., \u0026amp; Akers, M. D. (2007). A review of bankruptcy prediction studies: 1930 to present. \u003cem\u003eJournal of Financial education\u003c/em\u003e, 1-42.\u003c/li\u003e\n\u003cli\u003eBellovary, J. L., Giacomino, D. E., \u0026amp; Akers, M. D. (2007). A review of bankruptcy prediction studies: 1930 to present. \u003cem\u003eJournal of Financial education\u003c/em\u003e, 1-42.\u003c/li\u003e\n\u003cli\u003eBent\u0026eacute;jac, C., Cs\u0026ouml;rgő, A., \u0026amp; Mart\u0026iacute;nez-Mu\u0026ntilde;oz, G. (2021). A comparative analysis of gradient boosting algorithms. \u003cem\u003eArtificial Intelligence Review\u003c/em\u003e, \u003cem\u003e54\u003c/em\u003e, 1937-1967.\u003c/li\u003e\n\u003cli\u003eBiau, G. (2012). Analysis of a random forests model. \u003cem\u003eThe Journal of Machine Learning Research\u003c/em\u003e, \u003cem\u003e13\u003c/em\u003e, 1063-1095.\u003c/li\u003e\n\u003cli\u003eBragoli, D., Ferretti, C., Ganugi, P., Marseguerra, G., Mezzogori, D., \u0026amp; Zammori, F. (2022). Machine-learning models for bankruptcy prediction: do industrial variables matter?. \u003cem\u003eSpatial Economic Analysis\u003c/em\u003e, \u003cem\u003e17\u003c/em\u003e(2), 156-177.\u003c/li\u003e\n\u003cli\u003eBreiman, L. (2001). Random forests. Machine Learning, 45(1), 5\u0026ndash;32.\u003c/li\u003e\n\u003cli\u003eChaudhuri, A., \u0026amp; De, K. (2011). Fuzzy support vector machine for bankruptcy prediction. \u003cem\u003eApplied Soft Computing\u003c/em\u003e, \u003cem\u003e11\u003c/em\u003e(2), 2472-2486.\u003c/li\u003e\n\u003cli\u003eChen, T., \u0026amp; Guestrin, C. (2016). Xgboost: A scalable tree boosting system. In \u003cem\u003eProceedings of the 22nd acm sigkdd international conference on knowledge discovery and data mining\u003c/em\u003e (pp. 785-794).\u003c/li\u003e\n\u003cli\u003eDu Jardin, P. (2010). Predicting bankruptcy using neural networks and other classification methods: The influence of variable selection techniques on model accuracy. Neurocomputing, 73(10-12):2047\u0026ndash;2060.\u003c/li\u003e\n\u003cli\u003eDu Jardin, P. (2015). Bankruptcy prediction using terminal failure processes. \u003cem\u003eEuropean Journal of Operational Research\u003c/em\u003e, \u003cem\u003e242\u003c/em\u003e(1), 286-303.\u003c/li\u003e\n\u003cli\u003eDu\u0026eacute;nez-Guzm\u0026aacute;n, E. A., \u0026amp; Vose, M. D. (2013). No free lunch and benchmarks. \u003cem\u003eEvolutionary Computation, 21\u003c/em\u003e(2), 293-312.\u003c/li\u003e\n\u003cli\u003eEdmister, R. O. (1972). An empirical test of financial ratio analysis for small business failure prediction. \u003cem\u003eJournal of Financial and Quantitative analysis\u003c/em\u003e, \u003cem\u003e7\u003c/em\u003e(2), 1477-1493.\u003c/li\u003e\n\u003cli\u003eHand, D., Mannila, H., \u0026amp; Smyth, P. (2001). Principles of data mining. Cambridge, MA: MIT Press.\u003c/li\u003e\n\u003cli\u003eHuang, Z., Chen, H., Hsu, C. J., Chen, W. H., \u0026amp; Wu, S. (2004). Credit rating analysis with support vector machines and neural networks: a market comparative study. \u003cem\u003eDecision support systems\u003c/em\u003e, \u003cem\u003e37\u003c/em\u003e(4), 543-558.\u003c/li\u003e\n\u003cli\u003eImandoust, S. B., \u0026amp; Bolandraftar, M. (2013). Application of k-nearest neighbor (knn) approach for predicting economic events: Theoretical background. \u003cem\u003eInternational journal of engineering research and applications\u003c/em\u003e, \u003cem\u003e3\u003c/em\u003e(5), 605-610.\u003c/li\u003e\n\u003cli\u003eKasgari, Ahmad Ahmadpour, Seyyed Hasan Salehnezhad, and Fatemeh Ebadi. \u0026quot;The bankruptcy prediction by neural networks and logistic regression.\u0026quot; \u003cem\u003eInternational Journal of Academic Research in Accounting, Finance and Management Sciences\u003c/em\u003e 3.4 (2013): 146-152.\u003c/li\u003e\n\u003cli\u003eLiang, D., Lu, C. C., Tsai, C. F., \u0026amp; Shih, G. A. (2016). Financial ratios and corporate governance indicators in bankruptcy prediction: A comprehensive study. \u003cem\u003eEuropean journal of operational research\u003c/em\u003e, \u003cem\u003e252\u003c/em\u003e(2), 561-572.\u003c/li\u003e\n\u003cli\u003eMai, F., Tian, S., Lee, C., \u0026amp; Ma, L. (2019). Deep learning models for bankruptcy prediction using textual disclosures. \u003cem\u003eEuropean journal of operational research\u003c/em\u003e, \u003cem\u003e274\u003c/em\u003e(2), 743-758.\u003c/li\u003e\n\u003cli\u003eMin, J. H., \u0026amp; Lee, Y. C. (2005). Bankruptcy prediction using support vector machine with optimal choice of kernel function parameters. \u003cem\u003eExpert systems with applications\u003c/em\u003e, \u003cem\u003e28\u003c/em\u003e(4), 603-614.\u003c/li\u003e\n\u003cli\u003eMaione, C., Batista, B.L., Campiglia, A.D., Barbosa, F., Jr, \u0026amp; Barbosa, R.M. (2016). Classification of geographic origin of rice by data mining and inductively coupled plasma mass spectrometry. \u003cem\u003eComputers and Electronics in Agriculture\u003c/em\u003e, 121, 101\u0026ndash;107.\u003c/li\u003e\n\u003cli\u003eMuslim, M. A., \u0026amp; Dasril, Y. (2021). Company bankruptcy prediction framework based on the most influential features using XGBoost and stacking ensemble learning. \u003cem\u003eInternational Journal of Electrical and Computer Engineering (IJECE)\u003c/em\u003e, \u003cem\u003e11\u003c/em\u003e(6), 5549-5557.\u003c/li\u003e\n\u003cli\u003eNarvekar, A., \u0026amp; Guha, D. (2021). Bankruptcy prediction using machine learning and an application to the case of the COVID-19 recession. \u003cem\u003eData Science in Finance and Economics\u003c/em\u003e, \u003cem\u003e1\u003c/em\u003e(2), 180-195.\u003c/li\u003e\n\u003cli\u003eOdom, M. D., \u0026amp; Sharda, R. (1990). A neural network model for bankruptcy prediction. In \u003cem\u003e1990 IJCNN International Joint Conference on neural networks\u003c/em\u003e (pp. 163-168). IEEE.\u003c/li\u003e\n\u003cli\u003eOdom, M. D., \u0026amp; Sharda, R. (1990). A neural network model for bankruptcy prediction. In \u003cem\u003e1990 IJCNN International Joint Conference on neural networks\u003c/em\u003e (pp. 163-168). IEEE. \u003c/li\u003e\n\u003cli\u003eOhlson, J. A. (1980). Financial ratios and the probabilistic prediction of bankruptcy. \u003cem\u003eJournal of accounting research\u003c/em\u003e, 109-131.\u003c/li\u003e\n\u003cli\u003eOlson, D. L., Delen, D., \u0026amp; Meng, Y. (2012). Comparative analysis of data mining methods for bankruptcy prediction. \u003cem\u003eDecision Support Systems\u003c/em\u003e, \u003cem\u003e52\u003c/em\u003e(2), 464-473.\u003c/li\u003e\n\u003cli\u003eOlson, D. L., Delen, D., and Meng, Y. (2012). Comparative analysis of data mining methods for bankruptcy prediction. Decision Support Systems, 52(2):464\u0026ndash;473.\u003c/li\u003e\n\u003cli\u003ePerboli, G., \u0026amp; Arabnezhad, E. (2021). A Machine Learning-based DSS for mid and long-term company crisis prediction. \u003cem\u003eExpert Systems with Applications\u003c/em\u003e, \u003cem\u003e174\u003c/em\u003e, 114758.\u003c/li\u003e\n\u003cli\u003eRamraj, S., Uzir, N., Sunil, R., \u0026amp; Banerjee, S. (2016). Experimenting XGBoost algorithm for prediction and classification of different datasets. \u003cem\u003eInternational Journal of Control Theory and Applications\u003c/em\u003e, \u003cem\u003e9\u003c/em\u003e(40), 651-662.\u003c/li\u003e\n\u003cli\u003eSandri, M., \u0026amp; Zuccolotto, P. (2008). A bias correction algorithm for the Gini variable importance measure in classifcation trees. \u003cem\u003eJournal of Computational and Graphical Statistics, 17\u003c/em\u003e(3), 611\u0026ndash;628.\u003c/li\u003e\n\u003cli\u003eShrivastava, S., Jeyanthi, P. M., \u0026amp; Singh, S. (2020). Failure prediction of Indian Banks using SMOTE, Lasso regression, bagging and boosting. \u003cem\u003eCogent Economics \u0026amp; Finance, 8\u003c/em\u003e(1), 1729569.\u003c/li\u003e\n\u003cli\u003eShetty, S., Musa, M., \u0026amp; Br\u0026eacute;dart, X. (2022). Bankruptcy Prediction Using Machine Learning Techniques. \u003cem\u003eJournal of Risk and Financial Management\u003c/em\u003e, \u003cem\u003e15\u003c/em\u003e(1), 35.\u003c/li\u003e\n\u003cli\u003eSmith, M., \u0026amp; Alvarez, F. (2021). Predicting firm-level bankruptcy in the Spanish economy using extreme gradient boosting. \u003cem\u003eComputational Economics\u003c/em\u003e, 1-33.\u003c/li\u003e\n\u003cli\u003eSon, H., Hyun, C., Phan, D., \u0026amp; Hwang, H. J. (2019). Data analytic approach for bankruptcy prediction. \u003cem\u003eExpert Systems with Applications\u003c/em\u003e, \u003cem\u003e138\u003c/em\u003e, 112816.\u003c/li\u003e\n\u003cli\u003eTian, S. and Yu, Y. (2017). Financial ratios and bankruptcy predictions: An international evidence. \u003cem\u003eInternational Review of Economics \u0026amp; Finance,\u003c/em\u003e 51:510\u0026ndash;526.\u003c/li\u003e\n\u003cli\u003eTsai, C. F. (2008). Financial decision support using neural networks and support vector machines. \u003cem\u003eExpert Systems\u003c/em\u003e, \u003cem\u003e25\u003c/em\u003e(4), 380-393.\u003c/li\u003e\n\u003cli\u003eTsai, C. F., Hsu, Y. F., \u0026amp; Yen, D. C. (2014). A comparative study of classifier ensembles for bankruptcy prediction. \u003cem\u003eApplied Soft Computing\u003c/em\u003e, \u003cem\u003e24\u003c/em\u003e, 977-984.\u003c/li\u003e\n\u003cli\u003eTseng, F. M., \u0026amp; Hu, Y. C. (2010). Comparing four bankruptcy prediction models: Logit, quadratic interval logit, neural and fuzzy neural networks. \u003cem\u003eExpert systems with applications\u003c/em\u003e, \u003cem\u003e37\u003c/em\u003e(3), 1846-1853.\u003c/li\u003e\n\u003cli\u003eYeh, C.C., Chi, D.J., \u0026amp; Lin, Y.R. (2014). Going-concern prediction using hybrid random forests and rough set approach. \u003cem\u003eInformation Sciences, 254\u003c/em\u003e, 98-110.\u003c/li\u003e\n\u003cli\u003eWang, G., Ma, J., \u0026amp; Yang, S. (2014). An improved boosting based on feature selection for corporate bankruptcy prediction. \u003cem\u003eExpert Systems with Applications\u003c/em\u003e, \u003cem\u003e41\u003c/em\u003e(5), 2353-2361.\u003c/li\u003e\n\u003cli\u003eWang, G., Ma, J., Huang, L., \u0026amp; Xu, K. (2012). Two credit scoring models based on dual strategy ensemble trees. \u003cem\u003eKnowledge-Based Systems\u003c/em\u003e, \u003cem\u003e26\u003c/em\u003e, 61-68.\u003c/li\u003e\n\u003cli\u003eWilson, R. L., \u0026amp; Sharda, R. (1994). Bankruptcy prediction using neural networks. \u003cem\u003eDecision support systems\u003c/em\u003e, \u003cem\u003e11\u003c/em\u003e(5), 545-557.\u003c/li\u003e\n\u003cli\u003eWhitrow, C., Hand, D.J., Juszczak, P., Weston, D. \u0026amp; Adam, N.M. (2009). Transaction aggregation as a strategy for credit card fraud detection. \u003cem\u003eData Mining and Knowledge Discovery, 18\u003c/em\u003e, 30-55.\u003c/li\u003e\n\u003cli\u003eXie, E., Li, X., Ngai, E. \u0026amp; Ying, W. (2009). Customer churn prediction using improved balanced random forest. \u003cem\u003eExpert Systems with Applications, 36\u003c/em\u003e, 5445-5449.\u003c/li\u003e\n\u003cli\u003eZhang, W. (2017). Machine learning approaches to predicting company bankruptcy. \u003cem\u003eJournal of Financial Risk Management\u003c/em\u003e, \u003cem\u003e6\u003c/em\u003e(04), 364.\u003c/li\u003e\n\u003cli\u003eZhao, H., Sinha, A. P., \u0026amp; Ge, W. (2009). Effects of feature construction on classification performance: An empirical study in bank failure prediction. \u003cem\u003eExpert Systems with Applications\u003c/em\u003e, \u003cem\u003e36\u003c/em\u003e(2), 2633\u0026ndash;2644.\u003c/li\u003e\n\u003cli\u003eZorič\u0026aacute;k, M., Gnip, P., Drot\u0026aacute;r, P., \u0026amp; Gazda, V. (2020). Bankruptcy prediction for small-and medium-sized companies using severely imbalanced datasets. \u003cem\u003eEconomic Modelling\u003c/em\u003e, \u003cem\u003e84\u003c/em\u003e, 165-176.\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"bankruptcy prediction, financial ratios, machine learning, logistic regression, emerging country","lastPublishedDoi":"10.21203/rs.3.rs-5693739/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-5693739/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eThis paper investigates the optimal approach for predicting corporate bankruptcy risk in Vietnam, utilizing a unique dataset of listed firms from 2010 to 2021 based on financial ratios. The results confirm that machine learning models significantly outperform traditional logistic regression, with XGBoost and Random Forest demonstrating superior predictive power compared to K-Nearest Neighbor and logistic regression across both one-year and two-year forecast horizons. The study also contributes methodologically by incorporating additional evaluation metrics including F1 Score, AUC-ROC, Brier Score, and Log-loss to assess classification and probability prediction performance more comprehensively. Model performance is further enhanced through hyperparameter tuning, class imbalance adjustment, and grid search optimization. Empirical findings highlight the importance of all six financial indicator groups, with specific ratios such as Total Assets Turnover, Current Assets to Sales, Current Assets to Total Assets, Equity to Total Assets, and Total Equity Growth playing a critical role in predicting corporate failure. These indicators emphasize the importance of liquidity management, asset efficiency, and equity growth in determining a firm’s financial resilience. Overall, this study not only enhances forecasting accuracy through advanced modeling but also provides valuable insights for stakeholders, particularly financial institutions, investors, and corporate managers supporting more informed decision-making and proactive risk management in Vietnam’s dynamic and evolving business environment.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eJEL Codes: G33, G34, M10.\u003c/strong\u003e\u003c/p\u003e","manuscriptTitle":"Financial ratios as indicators in bankruptcy prediction: A comparative analysis of statistical and machine learning models","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-03-27 09:35:33","doi":"10.21203/rs.3.rs-5693739/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"64b07583-a087-4dad-991a-5341e596ef87","owner":[],"postedDate":"March 27th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2025-04-09T11:53:51+00:00","versionOfRecord":[],"versionCreatedAt":"2025-03-27 09:35:33","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-5693739","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-5693739","identity":"rs-5693739","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}