Exploring Machine learning Algorithms to Identify Determinants of Risky Behavior among Pregnant Women 15–59 years in Eastern African countries using the Demographic and Health Survey data

Exploring Machine learning Algorithms to Identify Determinants of Risky Behavior among Pregnant Women 15–59 years in Eastern African countries using the Demographic and Health Survey data | 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 Exploring Machine learning Algorithms to Identify Determinants of Risky Behavior among Pregnant Women 15–59 years in Eastern African countries using the Demographic and Health Survey data Halid Worku Jemil, Sonia Worku Semayneh, Altaseb Beyene Kassaw, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7364399/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 15 You are reading this latest preprint version Abstract Introduction Unhealthy consumption patterns of substances, sexual activity, and physical inactivity are key contributors to morbidity and mortality for pregnant women. However, there is a limited study on those risk behaviors and their determinants among pregnant women in East Africa. Therefore, this study aimed to determine risky behaviors and their determinants among pregnant women in East Africa by using data from the DHS using machine learning algorithms. Methods This study utilized DHS data from 2012–2022 in 12 East African countries. Data was analyzed using Python version 3.7 and R version 4.3.3 for data preprocessing, modeling, and statistical analysis. Model performance was evaluated using accuracy and Area Under the Curve (AUC). Finally, the SHAP was applied in Python to further explore and interpret the predictors of risky behaviors among pregnant women aged 15–59 years old. Results In this study, the Light Gradient Boosting Machine model achieved an accuracy of 95.88% and an AUC score of 0.991. The SHapley Additive exPlanations analysis revealed that pregnant women who lived in rural areas, women with poor wealth income, women with middle wealth income, women whose husbands had primary education, and women not exposed to media increased risky behavior. Whereas women who were employed, women’s utilized ANC services, and women aged 25–36 lower likelihood of risky behaviors. Conclusion The Light GBM was the best-performing model for identifying determinants of risky behaviors among pregnant women in Eastern African countries. Interventions should focus on promoting and strengthening women’s ANC accessibility, improving husbands’ education, expanding media use, and economic empowerment for women to reduce the burden of risky behaviors. Machine learning Risky behavior Light GBM pregnant women Shapley Additive Explanation (SHAP) Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 1. Introduction Health-related risky behavior encompasses actions or personal traits that destroy health and well-being both in the short and long term. These behaviors, which include activities such as smoking, excessive alcohol use, engaging in harmful practices, physical inactivity, and engaging in risky sexual activity, are destructive health behaviors that may lead to poor health care ( 1 , 2 ). At a global level, the majority of the burden of disease shifts to non-communicable diseases (NCDs) ( 3 , 4 ). There are about 185 million drug users, 1.3 billion smokers, and 2 billion alcohol drinkers from those 9.8% of pregnant women consume alcohol, with higher prevalence in Europe (25.2%) and certain low- and middle-income countries (LMICs)( 5 ) approximately 5.7% of pregnant women use tobacco( 6 ). Additionally, 5.4% and 3.7% of the global burden of disease are related to alcohol and tobacco use respectively ( 7 ). The prevalence of substance use in a combined way was 80% globally ( 8 , 9 ). Similarly, Physical inactivity during pregnancy affects 28% of women( 10 ). Collectively 7.2% of all deaths and 7.6% of cardiovascular deaths are linked to physical inactivity ( 11 ). In SSA, chewing khat among pregnant women ranges from 15–30%( 6 ). A study showed that substance use in Sub-Saharan Africa accounts to 41.6% and in Central Africa (55.5%) from those the most commonly used substances are caffeine products (41.2%), alcohol (32.8%), tobacco (23.5%), and khat (22%) ( 12 ). In Eastern African countries, substance use accounts for 43.70% ( 13 ), and khat users in African countries account for 22% ( 14 ), Comoros (23.90%), Uganda (65.60%), Ethiopia (68.70%), and Mozambique (76.70%) ( 15 ). Additionally, behaviors such as smoking, excessive alcohol consumption, and physical inactivity contribute significantly to the burden of disease and morbidity ( 15 , 16 ). Risk behaviors have a great impact on women's mental, psychological, social, and physical health. The effect of this high prevalence of health risks leads to high premature death related to non-communicable diseases (NCDs) such as cardiovascular diseases, diabetes, cancer, high blood pressure, and respiratory problems associated with harmful alcohol consumption, poor dietary intake, cigarette use, and physical inactivity ( 17 – 19 ). Similarly, risky behaviors contribute significantly to both communicable and non-communicable diseases ( 13 , 20 , 21 ). Moreover, risky behaviors among pregnant women have different health consequences, such as increased risks of chronic diseases, mortality, and disability, and a negative impact on physical and mental health ( 22 – 24 ). The WHO and CDC are conducting a Global School Health Survey to identify behavioral risks and focus on promoting physical activity, healthy diets, and regulating unhealthy food marketing to children ( 25 ). Moreover, many interventions are conducted to get rid of risky behaviors. However, they targeted only a single behavior, while few have addressed multiple behaviors to reduce risky behaviors, although the effects are often small and may not be sustained long-term to reduce public health burdens ( 26 – 32 ). Therefore, it is crucial to develop and evaluate appropriate interventions that target people with risky behaviors to help them refrain from engaging in risky behaviors. Existing efforts often focus on single issues, such as smoking or unhealthy diets, rather than tackling interrelated behaviors like physical inactivity, substance use, and poor nutrition simultaneously ( 30 , 33 , 34 ). Based on an extensive literature search, no studies have been conducted on machine learning models to identify risky behavior among pregnant women 15–59 years old. Additionally, no such studies have been conducted in Eastern African countries. Furthermore, most of the existing literature relied on traditional statistical models, limited geographic coverage, inability to assess the combined effects of multiple risk behaviors, and institution-based data records. Hence, to handle these issues, a machine learning approach could help to identify factors associated with risky behavior in a comprehensive way ( 35 ). Furthermore, because risky behaviors are complex and multifaceted, traditional statistical models like logistic regression and linear regression usually fall short in capturing nonlinear patterns and interactions when used to support early health prevention strategies. ( 36 – 38 ). Machine learning, on the other hand, can find new information and work with high-dimensional data. It can also give useful explanations through tools like SHAP( 39 ). 2. Methods 2.1. Data source and study design This study employed a population-based cross-sectional design utilizing secondary data extracted from the DHS conducted in 12 East African countries from 2012 to 2022. East Africa is a geographically diverse region extending from the Horn of Africa to parts of Southern Africa. This study used standard DHS data from 12 countries (Burundi (2016-17), Ethiopia (2016), Madagascar (2021), Comoros (2012), Rwanda (2019-20), Tanzania (2022), Mozambique (2022-23), Zimbabwe (2015), Kenya (2022), and Zambia (2015). The 12 nations were chosen based on the variables of interest being available in the respective databases (S1). 2.2. Population All reproductive women 15–59 years residing in 12 Eastern African countries were the source population. Whereas all pregnant women 15–59 years living in the 12 Eastern African countries and present in the household during the enumeration period were the study population. 2.3. Inclusion and exclusion criteria All pregnant women aged 15–59 years who had complete and valid responses for at least one of the risky behavior indicators in the DHS data in 12 Eastern African countries were included in the study. Whereas non-pregnant women aged 15–59 years were excluded from the study. 2.4 Sample size determination and sampling procedures A total of 199,083 reproductive women were included; of those, 184,746 women were not pregnant during the enumeration period. Finally, this study utilized a sample of 14,337 pregnant women from 12 Eastern African countries, who were selected according to the detailed flow chart for study participants, as presented in (S2). 2.5. Data collection DHS uses a standardized and validated questionnaire. It used a two-stage stratified sampling technique to select representative study participants. To begin with, the Enumeration Areas (EAs) were chosen using a probability method that was aligned with the size of each area, making sure the selection was done independently in every sampling group. In the next phase, homes were selected in a systematic way. The key demographic and health indicators were collected in each DHS ( 40 ). For this study, Individual Record (IR) datasets from the DHS from 2012 to 2022 were utilized. All the detailed information for the survey (such as the sampling method, the determination of the sample size, and the data collection procedure) is available in Demographic and Health Survey reports from the Measure DHS program website https://www.dhsprogram.com . 2.6 Study variables The outcome variable for the study was risky behavior. 2.7. Operational Definition Risky behavior was dichotomized into two categories by merging two variables together. The women were considered to be engaging in risky behavior if they exhibited at least one of the following behaviors: using tobacco products (smoke cigarette, smoke pipe full of tobacco, snuffs by noise, snuffs by mouth, smoke kreteks, smoke cigar or cheroots, smoke water pipe, chewing tobacco, chewing betel liquid with tobacco, and smoke other local substance), and engaging in unsafe sexual activity (condom used during last sex from partners, number of sexual partner excluding spouse, ever been forced to unwanted sexual activity by partner). Whereas, the women were considered not engaged in risk behaviors if they exhibited none of the above behaviors ( 41 , 42 ). 2.8. Data quality management This study employed on utilizing high-quality data to which guarantee the validity and reliability of the model predictive performance. Some of the important parts of data preprocessing that were talked about were completeness, accuracy, uniqueness, timeliness, and consistency. For example, Stata made sure that the data was complete by filling in missing values before any more analysis. To improve the accuracy of the data and make sure the results were reliable, outliers in the dataset were found and removed using the right statistical methods. 2.9. Data management and analysis This study used a machine learning approach based on Yufeng Guo’s 7 steps of ML and the frameworks of a previous study. The seven steps employed in the management and analysis include data collection, data preprocessing, model selection, model training, model evaluation, hyperparameter tuning, and making predictions ( 43 ). Python version 3.7 on Jupyter Notebook and R were used for data preprocessing, modeling, and statistical analysis. Data preprocessing The process of machine learning begins with data pre-processing, which involves modifying or encoding the data to make it suitable for computer interpretation (84). In machine learning workflow, this study was employed a continuous improvement process for models. This process included selecting and engineering relevant features, splitting the data, model training, and model evaluation, model optimization, choosing the top performer model, and deploying the selected model for prediction. Through an iterative approach, study was refining our models. Data cleaning Missing value in the dataset was inspected amd imputed by using the K-nearest neighbors (KNN) imputation was used to fill in the missing values. The KNN approach was specifically used to impute variables with less than 3% missing data, taking into account both mean and median values ( 44 ). Variables with more than 3% missing data were excluded from the analysis.Furthermore, multicollinearity was evaluated by using the correlation coefficient matrix; high multicollinearity was defined as having a correlation coefficient more than 0.8, hence the independent variables were not interfere with each other( 45 ). Dimensionality reduction Based on shadow features importance for comparison, Boruta iteratively confirms or disproves attributes based on statistical evidence, ensuring that only significant predictors are retained. The Boruta algorithm was employed in this study as a dimensionality reduction technique to enhance model performance and reduce data complexity. High-dimensional datasets often contain repetitive or superfluous features, which can lower model accuracy and increase computational load. The Boruta algorithm was performed on R v4.33, then the feature importance was visualized using Boruta feature importance using boxplots. Non-important variables were rejected by the algorithm; and dimensionality reduction involves decreasing the number of input features to increased model efficiency, improve model performance( 46 ). Feature engineering Feature engineering including the process of identifying, categorizing, acquiring, and encoding the most relevant characteristics from the available datasets to build machine learning models that are more accurate ( 47 ). This study employed one-hot encoding for nominal categorical variables and label encoding for ordinal categorical variables to encode the data ( 48 ). Data balancing To address the issue of data balancing for the outcome variable, the SMOTEENN (Synthetic Minority Over-Sampling Technique with the Edited Nearest Neighbor ) method was used. The SMOTEENN first generates additional synthetic samples for the minority class by including the existing samples within the feature space which helps to reducing the risk of overfitting. Model selection In our study, the dependent variable included a method for classifying risk behaviors. This study the appropriate classifier was selected based on the literatures that use before for the purpose of classification to make predictions of health risky behaviors. The study used the Scikit-learn package version 3.7 in Python running within Jupyter Notebook to implement the machine learning algorithm. The selection of these algorithms was based on their suitability for classification tasks and their compatibility with the characteristics of our dataset include Random Forest (RF), Logistic Regression (LR), LightGBM, Naïve bays, Artficial Neural Network (ANN) Decision Tree (DT), extreme gradient boosting (XGB), K-nearest neighbor (KNN), AdaBoost, support vector machine (SVM), Cat Boost was used in this study and implemented by python version 3.7 using each packages and on Jupiter notebook ( 49 – 51 ). Model building and evaluation To make good predictive models, you need to train the model. The model learned from the training and testing data split of 30/70%, 20/80%, and 90/10%. This different way of splitting the data lets us accurately measure how well each prediction model works. The 80/20% train test data worked best for our model. Several metrics including sensitivity, specificity, and AUC were used to evaluate the performance of the prediction models. For this study accuracy and AUC was implemented to measure to evaluate effective model evaluation with imbalanced data ( 52 ). Each of these metrics provides valuable understanding into various parts of model performance. Accuracy measures the overall correctness of the model’s predictions. It is the ratio of the number of correct predictions to total predictions. Precision quantifies the accuracy of positive predictions made by the model. It finds the ratio of true positive predictions to total predicted positives. Recall, also known as sensitivity or true positive rate, assesses the model’s ability to identify all positive instances it is effectively utilized in various computational models and machine learning applications ( 53 ). AUC is alternative measure other than accuracy to evaluating the performance of models, specifically in an unbalanced data sets, since it is threshold-independent and ability to provide a more overall performances for predictive models ( 54 ). Table 1 Evaluation metrics to identify the determinants of risky behavior among pregnant women, 2012–2022. Classes Predicted class YES NO Total Yes a (TN) b (FP) TN + FN No c (FN) d (TP) FN + TP Total TN + FN FP + TP TN + FN + FP + TP As illustrated in Table 1 two by two tables, the following formula were used to calculate. Accuracy = TP + TN /TP + TN + FP + FN Sensitivity = TP /TP + FN Specificity = TN /TN + FP Precision = TP/ TP + FP Utilizing the above metrics, the study comprehensively evaluated the performance of each predictive model in terms of overall correctness, accurate positive predictions, and identification of positive instances, balanced measure, and discriminatory ability. Hyperparameter tuning Hyperparameter is an external manipulation to the model whose value must be set by the user ( 55 ). The selected model was optimized with the best parameters by applying the Optuna with 10-fold cross validation on the specified search space with one hundred trials. Since these techniques is more efficient when we deal with complex models and larger dataset than other techniques by using python version 3.7 on Jupiter notebook. Model interpretability In this study, the interpretability of machine learning models was essential to ensure clinical relevance and support evidence-based decision-making. To address the complexity and black-box nature of advanced predictive models, SHapley Additive exPlanations (SHAP) were employed. SHAP, based on cooperative game theory, offers a unified and theoretically grounded method to explain individual predictions by fairly attributing the contribution of each feature to the model’s output ( 56 ). Among the various SHAP implementations, Tree SHAP was selected for its computational efficiency and exact calculations when applied to tree-based models. This was especially relevant since Light GBM was identified as the best-performing model in this study. SHAP played a central role in both feature selection and model interpretation. Features were levelled by considering their average absolute SHAP values, which handle to identify and the variability of the outcome variables. Moreover, a range of SHAP visualization techniques were used to improve both global and local model interpretability. The global feature importance plot provided an overall ranking of feature. Similarly, the beeswarm plot illustrated the direction and distribution of features across the population, and the waterfall plot offered detailed case-level explanations. These tools collectively facilitated the discovery of complex interactions and non-linear relationships between the predictors that are often overlooked using traditional statistical methods. In this study, the integration of SHAP provided a comprehensive analytical framework for understanding health risk behaviors among adult men in East Africa. SHAP enabled transparent and data-driven explanations of model predictions. The overall data preprocessing and analysis workflow is summarized and presented in Fig. 3 . 3. Result 3.1. Socio-demographic and Behavioral related characteristics of study participants Majority of the respondents have attained primary (47.7%), while 21.7% have no formal education. About 36.7% were categorized as rich, and 44.5% as poor. The majority reside in urban areas (74.8%). More than half of the respondents (54.5%) were currently working. About 16.56% of the study participant were engaged in some form of risky behaviors. Approximately 14.76% of participants were involved in unsafe sexual activity, while 2.16% of participants were using tobacco products. All the detailed descriptions of study participants are presented in Table 2 . Table 2 Socio-demographic and Behavioral related characteristics of pregnant women in Eastern African countries, 2012–2022 (n = 14337). Variable Categories Frequency Percentage Women’s age 15–24 5486 38.21 25–36 7550 32.7 >=37 1300 29.1 women’s education No education 2958 20.6 Primary 6846 47.7 Secondary & Higher 4535 31.6 Marital Status Single 1061 7.4 Divorced/separated 712 87.6 Married 12564 4.9 Women’s Occupation Not-working 6522 45.9 Working 7815 54.5 Husband’s Education No-Education 2211 17.6 Primary 5441 43.34 Secondary & Higher 4830 40 ANC Visits No 7190 51.93 Yes 7146 48 Wealth status Poor 6383 44.5 Middle 2688 18.7 Rich 5266 36.7 Household head sex Male 5699 74.97 Female 1902 25.03 Family size 1_4 7022 48.9 5– 8 5899 41.1 >=9 1416 9.88 Fuel type Not-Electricity 13635 95.1 Electricity 702 4.9 Media exposure No 5427 37.8 Yes 8910 62.1 Place of residence Urban 10736 74.88 Rural 3601 25.12 Risky behavior Yes 2374 16.56 No 11963 83.44 Unsafe sexual activity Yes 2115 14.75 No 12222 85.25 Tobacco use Yes 309 2.16 No 14028 97.84 3.2 Machine learning analysis results 3.2.1 Feature selection In this study, Boruta analysis identified several significant predictors of risky behavior. Features such as country of residence, age, sex of household head, mobile phone use, residence, marital status, and media exposure were consistently confirmed as important, as indicated by their high importance scores in the final boxplot output. These variables demonstrated a strong contribution to the model's predictive capacity. This guided the final modeling phase to focus on features with confirmed relevance, reducing noise and improving overall model performance. The figure below presents the final importance scores of the Boruta algorithm. Based on the Boruta algorithm, features that were represented with the green color and presented right to the Shadomax were included in the next analysis. However, marital status and family size were unimportant for model building and rejected by the algorithm as presented in Fig. 4 . 3.2.2 Data balancing In this study, the SMOTENN (Synthetic Minority Over-Sampling Technique combined with Edited Nearest Neighbors) algorithm was used to solve the problem of class imbalance in the outcome variable. The SMOTENN oversampling technique generated 9589 additional synthetic data for minority classes hence the data was changed from unbalanced to balanced distribution for both classes. The updated proportions greatly decreased the discrepancy while maintaining data integrity, even though the outcome did not produce a perfect 50:50 balance. This result was obtained by combining artificial oversampling of the minority class with ENN's elimination of noisy or ambiguous examples from the majority class. Consequently, the balanced dataset made it easier to train a more robust and equitable model, which improved performance as shown in Fig. 5 . 3.2.3 Model development and performance evaluation to predict risky behavior Based on the comprehensive evaluation of model performance metrics (Table 3 ), Light GBM demonstrates the highest performance among all models, achieving the top accuracy of 90.6% and the highest ROC AUC of 96.8%. XGBoost and Neural Network also performed competitively, with accuracies of 89.6% and 88.4%, and ROC AUC values of 0.9656 and 0.9497as illustrated in ROC curve results in (S5), respectively. In contrast, traditional models such as Logistic Regression and Naive Bayes showed relatively weaker performance, with accuracies below 70% and lower precision. Table 3 Evaluation matrices of balanced to identify determinants of risky behavior among pregnant women 15–59 years in Eastern African countries, 2012–2022. No ML model Accuracy score (%) Precision Sensitivity AUC score 1. Support Vector Machine (SVM) 88 0.8579 0.76425 0.7657 2. LightGBM 90 0.9018 0.8666 0.9680* 3. AdaBoost 89 0.8754 0.9301 0.8545 4. Cat boost 88.9 0.8984 0.8424 0.7545 5. XGBoost 89.5 0.8912 0.8515 0.9656 6. Neural Network (ANN) 88.4 0.8698 0.8455 0.9497 7. KNN 84.6 0.8349 0.738 0.9342 8. Random Forest (RF) 74.6 0.7855 0.5322 0.8347 9. Decision Tree (DT) 71.8 0.6465 0.699 0.8031 10. Naive Bayes 68 0.6417 0.5119 0.7248 11. Logistic Regression (LR) 66.2 0.6272 0.4494 0.7067 Maximum model performance * 3.2.4 Hyperparameter tuning for the best model After evaluating the performance of the model, Light GBM was the best-performing model in terms of both predictive accuracy and area under the curve. To further optimize Light GBM’s performance, we applied Bayesian Optimization using the Optuna framework after optimizing the default and optimized values are provided in Table 4 . Table 4 Default and optimal hyperparameters for Light GBM model to identify determinants of risky behavior among pregnant women 15–59 years in Eastern African countries, 2012–2022. No Hyperparameters Default value Optimum value 1 boosting_type ‘gdbt’ ‘dart’ 2 Objectives regression Binary 3 Metrics 12 Auc 4 n_estimators 100 300 5 num_leaves 31 100 6 learning_rates 0.1 0.3 7 max_depth -1 7 3.2.5 SHAP Summary Plot result A SHAP (SHapley Additive exPlanations) summary bar plot was created to comprehend the overall contribution of each feature in predicting risk behavior (Fig. 7 ). The mean absolute SHAP value for each of the top 9 most significant variables is displayed in this plot, indicating the average degree of influence each feature has on the model output for every individual, whether in a positive or negative direction. The mean SHAP value is shown on the x-axis, and the associated features are listed on the y-axis. Variable with more mean absolute SHAP values or variables appear on top of the bar reveal the most important predictors such as: women who lived in rural area (place residence_0), women’s with poor wealth income (wealth status_0), women’s with middle wealth income(wealth status_1), husbands education being primary(husbands education_1), women’s being employed(women’s occupation _1), women’s being not exposed to media (media_exposure_0), women’s being utilize ANC services (ANC visit_1), women’s age being 25–36(women’s age_1) are the top most important variables sorted in descending order from higher to lower as illustrated in Fig. 7 . 3.2.6 SHAP Beeswarm Plot result A SHAP (SHapley Additive exPlanations) beeswarm plot was used to provide a rich overview of how the variables impact the model’s predictions across all the data. The color denotes the feature value's magnitude (blue for low, red for high), and the x-axis represents the SHAP value, which quantifies the influence of a feature on the model output. The analysis was identified women who lived in rural area (place residence_0), women’s with poor wealth income (wealth status_0), women’s with middle wealth income (wealth status_1), husband’s education being primary (husbands education_1), and women’s being not exposed to media (media_exposure_0) increase the likelihood of risky behavior among pregnant women. Whereas women’s being employed (women’s occupation _1), women’s being utilized ANC services (ANC visit_1), and women’s age being 25–36 (women’s age_1) lowers the likelihood of risky behaviors among pregnant women, as illustrated in Fig. 8 . 4. Discussion This study aimed to identify determinants of risky behaviors among pregnant women in Eastern African countries using various machine learning models applied on the DHS data. For this purpose, 11 machine learning models were trained on the balanced training data through tenfold cross-validation. The performance of those 11 classifier models was compared by their classification accuracy and AUC score. During the first phase of predictive modeling on balanced training data, Light GBM performed better than other classifiers with an accuracy of 90.6% and a 0.968 AUC score. Hence, light GBM was the best predictive model and further analysis was performed after optimizing it for its optimal parameters. Light GBM demonstrated a strong ability to distinguish between individuals with and without risky behaviors. Similarly, a study conducted in China suggested that gradient boosting is the best model to predict risky sexual behaviors among university students ( 57 ). Based on the study finding that women living in rural areas had the highest probability of engaging in risky behaviors compared to urban. This finding is consistent with a study conducted in Virginia( 58 ), Myanmar ( 59 ) and Ethiopia ( 41 ). This might be due to the traditional lifestyles, lack exposed to health education, inadequate access nutrition, or insufficient prenatal care less regulated access to drugs in rural areas, cultural norms that promote alcohol consumption, and a lack of public health initiatives could be contributing to the rise in risky behavior ( 60 , 61 ). Women being from lower and middle wealth indices were more likely to engage in risky behaviors. This study is consistent with the study conducted in Canada ( 62 ), Denmark ( 63 ), and Pomerania ( 64 ). This could be because financial constraints may limit access to health-related information, wholesome food, and first-rate medical care. Women may also get stressed and frustrated due to financial hardship. Additionally, women with lower wealth levels may be more vulnerable to factors like limited access to services, financial stress, or living in lower socioeconomic status conditions ( 65 ). Moreover, paternal education being primary increases the probability of risky behaviors among pregnant women. This finding aligns with other studies conducted in Uganda ( 66 ) and Australia ( 67 ), showing that paternal education level significantly affects maternal health conditions. This might be due to male partner education being lower might be associated with the delay in intention to detect early pregnancy signs and symptoms, lower health care decision making, decreased support and counseling for pregnancy risks, and insufficient support for maternal healthful behavior ( 68 , 69 ). Furthermore, women who were not exposed to the media were more likely to engage in risky activities ( 70 ). This may be because the media is an important medium to transmitting essential public health messages, such as adequate dietary intake for pregnant mothers, ANC service utilization, environmental hygiene and sanitation, raising awareness among women on early pregnancy complications, and strengthening ANC care practices ( 71 ). Whereas women who are employed or working women decrease risky behaviors during pregnancy. This study was conducted in Australia ( 67 ) and Iran ( 72 ). This may be due to occupation can strengthen women's finances and increase their reliance on making a healthy lifestyle, reducing stress, including utilization of ANC care services on a timely basis, and promoting better healthful behaviors ( 67 ). Similarly, women who attend ANC care services were less likely to engage in risky behaviors. This finding is aligned with other studies conducted in Iran ( 73 ) and Pomerania ( 64 ). This could be due to that ANC visits providing opportunities for health education, early detection of pregnancy complications, empowering women’s to have adequate knowledge about proper nutrition, substance use and provide counseling service on nutrition and substance avoidance, and empowering women’s to make better health decision all of which contribute to reducing risky behaviors ( 74 , 75 ). Additionally, older women aged 25 and 36 were less likely to engage in risky behaviors. This age group is strongly associated to higher maturity and self-regulation as well as they have strong background for unhealthy patterns and better ability for impulse control and unwanted behaviors than younger women in handling, and physiological well-being for pregnancy. This finding is supported by studies conducted in Denmark ( 63 ), Ethiopia ( 76 ), Scotland ( 18 ), and South Africa ( 30 , 77 ). This could be because younger women (< 25 years) may lack experience, while older women may engage in less risky behaviors. This might be due to older women have a lower probability of engaging in risky behaviors like unsafe sex, transactional sex, use of substance use, or peer pressure influence ( 78 , 79 ). 5. Strengths and Limitations This study employed standardized measurements for measuring risky behaviors, which is internationally recognized and adopted by the WHO and other organizations widely used as a standardized tool for measuring risky behaviors in this study. Consequently, comparison of result across many studies will be easy. Similarly, the study used SHAP model interpretation techniques which enables us to interpret complex models accurately, which bases on an optimized and best-performing model, which increases model performance to accurately identify risky behaviors among pregnant women. However, the study design was cross-sectional it does not show a causal relationship between the variables (cause-and-effect), similarly, it also does not show long-term effects. Moreover, the secondary nature of the dataset makes it difficult to explore and find additional essential features to accurately predict and identify associated factors for risky behavior in compressively. 6. Conclusion This study aims to identify determinants of risky behavior among pregnant women aged 15–59 years in Eastern African countries. The light GBM notably showed an outstanding performance with an accurate score of 95.88% and an AUC score of 0.991. Similarly, the findings revealed that women who lived in rural areas, women with poor wealth income, women with middle wealth income, husbands' education being primary, and women not being exposed to media increase the likelihood of risky behavior among pregnant women. Whereas, women’s being employed, women utilized ANC services, women’s age being 25–36 lowers the likelihood of risky behaviors among pregnant women. Abbreviations ANC – Ante Natal Care ANN – Artificial neural Networks AUC - Area under the Curve CV - Cross Validation CDC – Communicable Disease Control DHS - Demographic and Health Survey DT - Decision Tree EA - East Africa FN - False Negative FP – False Positive GBM - Light Gradient Boosting Machine IR – Individual Record KNN - K-nearest neighbor LR - Logistic Regression ML - Machine Learning NCD – Non-communicable Disease RF - Random Forest ROC - Receiving Operating Characteristics SHAP - Shapely Additive Explanation SSA - Sub-Saharan Africa SVM - Support Vector Machine SMOTEENN - Synthetic Minority Oversampling Techniques with the Edited Nearest Neighbor TN - True Negative TP – True Positive UNICEF - United Nations International Children’s Emergency Fund WHO - World Health Organization XGB - Extreme Gradient Boosting Declarations Authors’ Contributions H.W.J., did the conceptualization and investigation, A.B.K., S.W.S., and H.W.J. did data curation, H.W.J. and S.W.S. wrote the methodology. H.W.J., carried out the analysis. SW.S. H.W.J. wrote the original draft, A.E.B., A.B.K., and S.W.S. wrote and edited the final manuscript text. All authors subsequently revised the final manuscript. Conflict of interest The authors declare that they have no conflict of interest. Availability of data and materials All relevant data are within the manuscript and its Supporting Information files. Consent for publication Not applicable Ethical approval and consent from participants Data was obtained after granted permission through proper registration to access the data from the DHS website https://www.dhsprogram.com. The data for this study was downloaded after full authorization and approval, and consent from the DHS committee with a formal letter, which is attached at the end of the document. Funding No funding was obtained for this study. Acknowledgements The authors acknowledge to Measure of DHS program committee for authorizing us to use the data sets. References Åsberg K, Lundgren O, Henriksson H, Henriksson P, Bendtsen P, Löf M, et al. Multiple lifestyle behaviour mHealth intervention targeting Swedish college and university students: protocol for the Buddy randomised factorial trial. BMJ Open. 2021;11(12):e051044. Duffy KA, Green PA, Chartrand TL. Mimicry and modeling of health (-risk) behaviors: How others impact our health (-risk) behaviors without our awareness. Journal of Nonverbal Behavior. 2020;44(1):5-40. Bourne RR. Global burden of 369 diseases and injuries in 204 countries and territories, 1990–2019: a systematic analysis for the Global Burden of Disease Study 2019. 2023. Bhattacharya S, Heidler P, Varshney S. Incorporating neglected non-communicable diseases into the national health program—A review. Frontiers in public health. 2023;10:1093170. Alcohol WGSRo. Global status report on alcohol and health 2018: Alcohol, Drugs and Addictive Behaviors (ADA); 2018 [ CDC. CDC Morbidity and mortality weekley report2019. Degenhardt L, Charlson F, Ferrari A, Santomauro D, Erskine H, Mantilla-Herrara A, et al. The global burden of disease attributable to alcohol and drug use in 195 countries and territories, 1990–2016: a systematic analysis for the Global Burden of Disease Study 2016. The Lancet Psychiatry. 2018;5(12):987-1012. Glantz MD, Bharat C, Degenhardt L, Sampson NA, Scott KM, Lim CC, et al. The epidemiology of alcohol use disorders cross-nationally: Findings from the World Mental Health Surveys. Addictive behaviors. 2020;102:106128. Organization WH. Global status report on alcohol and health 2018: World Health Organization; 2018. WHO. WHO guidelines on physical activity and sedentary behaviour2020. Katzmarzyk PT, Friedenreich C, Shiroma EJ, Lee IM. Physical inactivity and non-communicable disease burden in low-income, middle-income and high-income countries. Br J Sports Med. 2022;56(2):101-6. Sreeramareddy CT, Pradhan PM, Sin S. Prevalence, distribution, and social determinants of tobacco use in 30 sub-Saharan African countries. BMC medicine. 2014;12:1-13. Fentaw KD, Fenta SM, Biresaw HB. Prevalence and Associated factors of Substance Use Male Population in East African countries: a multilevel analysis of recent demographic and health surveys from 2015 to 2019. Substance abuse: research and treatment. 2022;16:11782218221101011. . Park JH, Moon JH, Kim HJ, Kong MH, Oh YH. Sedentary lifestyle: overview of updated evidence of potential health risks. Korean journal of family medicine. 2020;41(6):365. Purba MAK, Thomson DRM, Henery DPM, Pearce DA, Henderson PM, Katikireddi PSV. Social media use and adolescent health-risk behaviours: A systematic review and meta-analysis. medRxiv. 2023:2023.02. 24.23286415. Bista B, Dhimal M, Bhattarai S, Neupane T, Xu YY, Pandey AR, et al. Prevalence of non-communicable diseases risk factors and their determinants: Results from STEPS survey 2019, Nepal. PloS one. 2021;16(7):e0253605. Olajide D, Eberth B, Ludbrook A. Analysis of multiple health risky behaviours and associated disease outcomes using Scottish linked hospitalisation data. Frontiers in public health. 2022;10:847938. Meader N, King K, Moe-Byrne T, Wright K, Graham H, Petticrew M, et al. A systematic review on the clustering and co-occurrence of multiple risk behaviours. BMC public health. 2016;16:1-9. Fast D, Bukusi D, Moyer E. The knife's edge: masculinities and precarity in East Africa. Social Science & Medicine. 2020;258:113097. Huijts T, Gkiouleka A, Reibling N, Thomson KH, Eikemo TA, Bambra C. Educational inequalities in risky health behaviours in 21 European countries: findings from the European social survey (2014) special module on the social determinants of health. The European Journal of Public Health. 2017;27(suppl_1):63-72. Okereke O. INFLUENCE OF PERSONALITY ON SEXUAL RISK BEHAVIOUR. Journal of Institutional Research, Big Data Analytics and Innovation. 2025;1(2):17-24. Fernandez A, Lozano A, Lee TK, Estrada Y, Messiah SE, Prado G. A family-based healthy lifestyle intervention: Crossover effects on substance use and sexual risk behaviors. Prevention Science. 2021;22:602-8. Gardner LA, Champion KE, Parmenter B, Grummitt L, Chapman C, Sunderland M, et al. Clustering of six key risk behaviors for chronic disease among adolescent females. International Journal of Environmental Research and Public Health. 2020;17(19):7211. who. Health education and promotion-Behavioural risk factors 2024 [Available from: (https://www.emro.who.int/health-education/health-risk-factors/behaviour-risk-factors.html). Silva CC, Presseau J, van Allen Z, Schenk PM, Moreto M, Dinsmore J, et al. Effectiveness of interventions for changing more than one behavior at a time to manage chronic conditions: A systematic review and meta-analysis. Annals of Behavioral Medicine. 2024;58(6):432-44. Bourke M, Patten RK, Dash S, Pascoe M, Craike M, Firth J, et al. The effect of interventions that target multiple modifiable health behaviors on symptoms of anxiety and depression in young people: a meta-analysis of randomized controlled trials. Journal of Adolescent Health. 2022;70(2):208-19. Tzelepis F, Mitchell A, Wilson L, Byrnes E, Haschek A, Leigh L, et al. The long-term effectiveness of internet-based interventions on multiple health risk behaviors: systematic review and robust variance estimation meta-analysis. Journal of Medical Internet Research. 2021;23(12):e23513. Tinner L, Palmer JC, Lloyd EC, Caldwell DM, MacArthur GJ, Dias K, et al. Individual-, family-and school-based interventions to prevent multiple risk behaviours relating to alcohol, tobacco and drug use in young people aged 8-25 years: a systematic review and meta-analysis. BMC Public Health. 2022;22(1):1111. Daniels KJ, Hoosen I, Pharaoh H. Health risk behaviour prevention/intervention programmes targeted at youth/adolescents engaging in risky behaviour—A scoping review. Journal of Public Health. 2024:1-10. Gómez-Gómez I, Moreno-Peral P, López T, Clavería A, Oliván B, Marti R, et al. Multiple risk behaviour intervention to prevent depression in primary care. British Journal of General Practice. 2020;70(suppl 1). Champion KE, Parmenter B, McGowan C, Spring B, Wafford QE, Gardner LA, et al. Effectiveness of school-based eHealth interventions to prevent multiple lifestyle risk behaviours among adolescents: a systematic review and meta-analysis. The Lancet Digital Health. 2019;1(5):e206-e21. Boyle MA, Hoffmann AN, Horn J, Badger S, Gaskill LA. Approaches for Treating Multiply Controlled Problem Behavior. Behavior Analysis in Practice. 2024;17(1):53-69. Noh S, Choo J. Gender-Specific Clustering of Lifestyle Behaviors and Its Impacts on Cardiovascular Health. Journal of Cardiovascular Nursing. 2024;39(4):E115-E25. Uddin S, Khan A, Hossain ME, Moni MA. Comparing different supervised machine learning algorithms for disease prediction. BMC medical informatics and decision making. 2019;19(1):1-16. Pasha SJ, Mohamed ES. Novel Feature Reduction (NFR) model with machine learning and data mining algorithms for effective disease risk prediction. Ieee Access. 2020;8:184087-108. Naveenkumar S, Kirubhakaran R, Jeeva G, Shobana M, Sangeetha K. Smart health prediction using machine learning. International Research Journal on Advanced Science Hub (IRJASH). 2021;3(3):124-8. Morgenstern JD, Buajitti E, O’Neill M, Piggott T, Goel V, Fridman D, et al. Predicting population health with machine learning: a scoping review. BMJ open. 2020;10(10):e037860. Mateen BA, Liley J, Denniston AK, Holmes CC, Vollmer SJ. Improving the quality of machine learning in health applications and clinical research. Nature Machine Intelligence. 2020;2(10):554-6. Croft TN, Marshall AM, Allen CK, Arnold F, Assaf S, Balian S. Guide to DHS statistics. Rockville: ICF. 2018;645:292-303. Kebede SA, Weldesenbet AB, Tusa BS. Spatial Distribution and Factors Associated With Risky Health Behavior Among Adult Males Aged 15–59 Years in Ethiopia: Generalized Structural Equation Modeling. Frontiers in Psychiatry. 2022;13:688336. Brener ND, Collins JL. Co-occurrence of health-risk behaviors among adolescents in the United States. J Adolesc Health. 1998;22(3):209-13. Kumar M, Yadav P, Singh H, Arora A, editors. Efficient Methods to Improve the Performance of Supervised Learning Models. 2021 International Conference on Intelligent Technologies (CONIT); 2021: IEEE. Rawat S, Rawat A, Kumar D, Sabitha AS. Application of machine learning and data visualization techniques for decision support in the insurance sector. International Journal of Information Management Data Insights. 2021;1(2):100012. Choi S-H, Oh Y-K, Koo D, editors. Visualization Techniques for Outlier Data. 2020 International Conference on Artificial Intelligence in Information and Communication (ICAIIC); 2020: IEEE. Stańczyk U, Jain LC. Feature selection for data and pattern recognition: An introduction: Springer; 2015. Zheng A, Casari A. Feature engineering for machine learning: principles and techniques for data scientists: " O'Reilly Media, Inc."; 2018. Al-Shehari T, Alsowail RA. An insider data leakage detection using one-hot encoding, synthetic minority oversampling and machine learning techniques. Entropy. 2021;23(10):1258. Kim M, Goh G. A sparse empirical Bayes approach to high‐dimensional Gaussian process‐based varying coefficient models. Stat. 2024;13(2):e678. Talekar B, Agrawal S. A detailed review on decision tree and random forest. Biosci Biotechnol Res Commun. 2020;13(14):245-8. Zhang S. Challenges in KNN classification. IEEE Transactions on Knowledge and Data Engineering. 2021;34(10):4663-75. Szeghalmy S, Fazekas A. A comparative study of the use of stratified cross-validation and distribution-balanced stratified cross-validation in imbalanced learning. Sensors. 2023;23(4):2333. Chen H, Wang N, Du X, Mei K, Zhou Y, Cai G. Classification prediction of breast cancer based on machine learning. Computational intelligence and neuroscience. 2023;2023(1):6530719. Carrington AM, Manuel DG, Fieguth PW, Ramsay T, Osmani V, Wernly B, et al. Deep ROC analysis and AUC as balanced average accuracy, for improved classifier selection, audit and explanation. IEEE Transactions on Pattern Analysis and Machine Intelligence. 2022;45(1):329-41. Brownlee J. Data preparation for machine learning: data cleaning, feature selection, and data transforms in Python: Machine Learning Mastery; 2020. Merrick L, Taly A, editors. The explanation game: Explaining machine learning models using shapley values. Machine Learning and Knowledge Extraction: 4th IFIP TC 5, TC 12, WG 84, WG 89, WG 129 International Cross-Domain Conference, CD-MAKE 2020, Dublin, Ireland, August 25–28, 2020, Proceedings 4; 2020: Springer. Li X, Zhang H, Zhao S, Tang K. Predicting risky sexual behavior among college students through machine learning approaches: cross-sectional analysis of individual data from 1264 universities in 31 provinces in China. JMIR public health and surveillance. 2023;9(1):e41162. Pilehvari A, Chipoletti A, Krukowski R, Little M. Unveiling socioeconomic disparities in maternal smoking during pregnancy: a comprehensive analysis of rural and Appalachian areas in Virginia utilizing the multi-dimensional YOST index. BMC Pregnancy and Childbirth. 2024;24(1):828. Htet AS, Bjertness MB, Sherpa LY, Kjøllesdal MK, Oo WM, Meyer HE, et al. Urban-rural differences in the prevalence of non-communicable diseases risk factors among 25–74 years old citizens in Yangon Region, Myanmar: a cross sectional study. BMC Public Health. 2016;16(1):1225. Nyirenda HT, Mulenga D, Silitongo M, Nyirenda HB, Nyirenda T. Spatial distribution and correlates of smoking in Zambia. BMJ open. 2019;9(8):e030044. Koblinsky M, Anwar I, Mridha MK, Chowdhury ME, Botlero R. Reducing maternal mortality and improving maternal health: Bangladesh and MDG 5. J Health Popul Nutr. 2008;26(3):280-94. Dennis CL, Brennenstuhl S, Brown HK, Bell RC, Marini F, Birken CS. High-risk health behaviours of pregnancy-planning women and men: Is there a need for preconception care? Midwifery. 2022;106:103244. Raru TB, Ayana GM, Zakaria HF, Merga BT. Association of higher educational attainment on antenatal care utilization among pregnant women in east africa using Demographic and Health Surveys (DHS) from 2010 to 2018: a multilevel analysis. International journal of women's health. 2022:67-77. Lange AE, Mahlo-Nguyen J, Pierdant G, Allenberg H, Heckmann M, Ittermann T. Antenatal Care and Health Behavior of Pregnant Women-An Evaluation of the Survey of Neonates in Pomerania. Children (Basel). 2023;10(4). Cortes ML, Louzado JA, Oliveira MG, Bezerra VM, Mistro S, Medeiros DS, et al. Association between perceived stress and health-risk behaviours in workers. Psychol Health Med. 2022;27(4):746-60. Theuring S, Rubagumya K, Schumann H, Harms G, Rubaihayo J, Wanyenze R. Sexual Risk Behavior in HIV-Uninfected Pregnant Women in Western Uganda. Arch Sex Behav. 2021;50(7):3277-86. Sarofim M, Huilan X, Wen LM, Baur LA. Associations of Sociodemographic Factors With Clusters of Obesity-Risk Behaviours Among Pregnant Women in Australia. Aust N Z J Obstet Gynaecol. 2025. Murakami K, Hashimoto H. Associations of education and income with heavy drinking and problem drinking among men: evidence from a population-based study in Japan. BMC Public Health. 2019;19(1):420. Kebede SA, Weldesenbet AB, Tusa BS. Spatial Distribution and Factors Associated With Risky Health Behavior Among Adult Males Aged 15-59 Years in Ethiopia: Generalized Structural Equation Modeling. Front Psychiatry. 2022;13:688336. Davis AM, Wambach KA, Nelson EL, Odar C, Lillis T, McKinley A, et al. Health behavior change in pregnant women: a two-phase study. Telemed J E Health. 2014;20(12):1165-9. Ramadani RC, Ibrahim K, Mirwanti R, Maulana S, Jabareen R. Social media use, knowledge, attitudes, and risky sexual behavior of HIV transmission: A survey among boarding school adolescent students in Indonesia. Belitung Nurs J. 2024;10(3):322-31. Tafti SP, Azarshab A, Mahmoudian RA, Khayami R, Abadi RNS, Ahmadi-Simab S, et al. Predictors of tobacco use among pregnant women: a large-scale, retrospective study. BMC Public Health. 2024;24(1):2544. Mirzakhani K, Shoorab NJ, Akbari A, Khadivzadeh T. High-risk pregnant women's experiences of the receiving prenatal care in COVID-19 pandemic: a qualitative study. BMC Pregnancy Childbirth. 2022;22(1):363. Rockliffe L, Peters S, Heazell AEP, Smith DM. Factors influencing health behaviour change during pregnancy: a systematic review and meta-synthesis. Health Psychol Rev. 2021;15(4):613-32. Downe S, Finlayson K, Tunçalp Ö, Gülmezoglu AM. Provision and uptake of routine antenatal services: a qualitative evidence synthesis. Cochrane Database Syst Rev. 2019;6(6):Cd012392. Akalu TY, Baraki AG, Wolde HF, Lakew AM, Gonete KA. Factors affecting current khat chewing among male adults 15–59 years in Ethiopia, 2016: a multi-level analysis from Ethiopian Demographic Health Survey. BMC psychiatry. 2020;20:1-8. Rawlings G, Williams R, Clarke D, English C, Fitzsimons C, Holloway I, et al. Exploring adults’ experiences of sedentary behaviour and participation in non-workplace interventions designed to reduce sedentary behaviour: a thematic synthesis of qualitative studies. BMC public health. 2019;19:1-16. Shahrabi Farahani F, Khosrowabadi R, Jaafari G. Risk-taking Behavior Under the Effect of Emotional Stimuli Among Children and Adults. Basic Clin Neurosci. 2022;13(4):585-93. Konlan P, Ganle JK. Transactional sex and associated factors among young women in a tertiary institution in Northern Ghana: evidence from a cross-sectional survey. BMC Womens Health. 2025;25(1):298. Additional Declarations No competing interests reported. Supplementary Files S1.tif S1. Figure. Map of study area Eastern African Countries to identify the determinants of risky behavior among pregnant women, 2012-2022. S2.tif S2. Figure. Sample size selection flow chart to identify the determinants of risky behavior among pregnant women, 2012-2022. S5.tif S5. Figure. ROC curve analysis of selected machine learning algorithms to identify determinants of risky behavior among pregnant women 15–59 yearsin Eastern African countries, 2012-2022. S6.pdf S6. PDF. DHS Permission Letter Cite Share Download PDF Status: Under Review Version 1 posted Editorial decision: Revision requested 28 Sep, 2025 Reviews received at journal 27 Sep, 2025 Reviews received at journal 19 Sep, 2025 Reviewers agreed at journal 14 Sep, 2025 Reviews received at journal 13 Sep, 2025 Reviewers agreed at journal 09 Sep, 2025 Reviewers agreed at journal 07 Sep, 2025 Reviewers agreed at journal 07 Sep, 2025 Reviewers agreed at journal 05 Sep, 2025 Reviewers agreed at journal 05 Sep, 2025 Reviewers agreed at journal 05 Sep, 2025 Reviewers invited by journal 05 Sep, 2025 Editor assigned by journal 13 Aug, 2025 Submission checks completed at journal 13 Aug, 2025 First submitted to journal 13 Aug, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7364399","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":512307941,"identity":"4f4e2d1d-8ee1-4524-ac7c-fb6e2c1488fc","order_by":0,"name":"Halid Worku Jemil","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA+klEQVRIiWNgGAWjYBACAwbmhgM8QIYEA/PBBx8qgCxm5gYCWhhhWtiSDWecAWlhJKyFAaKFx0yatw0kRkCLOfvBxgNv22zyJWckGBvOnFcbzd8O1PKjYhtOLZY9iQ0H57alWc6WSEh88HHb8dwZhxkbGHvO3MbtsAOJDYd5tx02kJNIOGw4c9ux3AagFmbGNjxazj+EaUlsk+adcyx3PkEtN6C2SEsks0nzNtTkbiCs5SHQL//SDCR7njEbzjh2IHcjUMtBvH45n3z4w5szNgYSx/M/PvhQU5c77/zhgw9+VODWgg4Og8kDRKsHgjpSFI+CUTAKRsEIAQA1EWYE1s3jEwAAAABJRU5ErkJggg==","orcid":"","institution":"Wollo University","correspondingAuthor":true,"prefix":"","firstName":"Halid","middleName":"Worku","lastName":"Jemil","suffix":""},{"id":512307942,"identity":"ceb901d4-6c82-4c49-84b4-5967bea56aac","order_by":1,"name":"Sonia Worku Semayneh","email":"","orcid":"","institution":"Addis Ababa University College of Health Science, Tikur Anbessa specialized hospital, Addis Ababa University","correspondingAuthor":false,"prefix":"","firstName":"Sonia","middleName":"Worku","lastName":"Semayneh","suffix":""},{"id":512307943,"identity":"d71bd801-f3c7-446f-acde-b7a505530d96","order_by":2,"name":"Altaseb Beyene Kassaw","email":"","orcid":"","institution":"Wollo University","correspondingAuthor":false,"prefix":"","firstName":"Altaseb","middleName":"Beyene","lastName":"Kassaw","suffix":""},{"id":512307944,"identity":"5bde3be7-0eea-4344-909b-ca5cf7288c1c","order_by":3,"name":"Anmut Endalkachew Bezie","email":"","orcid":"","institution":"Wollo University","correspondingAuthor":false,"prefix":"","firstName":"Anmut","middleName":"Endalkachew","lastName":"Bezie","suffix":""}],"badges":[],"createdAt":"2025-08-13 11:23:12","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7364399/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7364399/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":91307962,"identity":"49646cf0-2ff2-4fbb-a409-a637da75632a","added_by":"auto","created_at":"2025-09-15 06:43:50","extension":"jpg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":108167,"visible":true,"origin":"","legend":"\u003cp\u003eFigure 3: Workflow for identify determinants of risky behavior among pregnant women 15–59 years in Eastern African Countries, 2012-2022.\u003c/p\u003e","description":"","filename":"3.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7364399/v1/78e9c4dabdf8e211e9a3b9f5.jpg"},{"id":91307963,"identity":"a934588a-1a16-4aa2-b745-4f98066b41aa","added_by":"auto","created_at":"2025-09-15 06:43:50","extension":"jpg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":91483,"visible":true,"origin":"","legend":"\u003cp\u003eFigure 4: Boruta feature importance plot to identify determinants of risky behavior among pregnant women 15–59 years in Eastern African countries, 2012-2022.\u003c/p\u003e","description":"","filename":"4.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7364399/v1/b9a74514a1b76d8a22a73111.jpg"},{"id":91309248,"identity":"ea0f2b00-be02-485a-913c-3c76d22a052d","added_by":"auto","created_at":"2025-09-15 06:59:50","extension":"jpg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":60616,"visible":true,"origin":"","legend":"\u003cp\u003eFigure 5: Distribution of risky behavior among pregnant women before and after data balancing using SMOTENN to identify determinants of risky behavior among pregnant women 15–59 years in Eastern African countries, 2012-2022.\u003c/p\u003e","description":"","filename":"5.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7364399/v1/bacb9ed6badb342ea5060153.jpg"},{"id":91307966,"identity":"1951e45b-c759-49c1-b341-d374cd538fb5","added_by":"auto","created_at":"2025-09-15 06:43:50","extension":"jpg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":86552,"visible":true,"origin":"","legend":"\u003cp\u003eFigure \u003cstrong\u003e7\u003c/strong\u003e: SHAP global importance plot to identify determinants of risky behavior among pregnant women 15–59 years in Eastern African countries, 2012-2022.\u003c/p\u003e","description":"","filename":"7.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7364399/v1/85a68465fba51f0e0e3a82c5.jpg"},{"id":91307974,"identity":"eb5a64e3-b921-4682-8c5c-81bd411f877b","added_by":"auto","created_at":"2025-09-15 06:43:50","extension":"jpg","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":98225,"visible":true,"origin":"","legend":"\u003cp\u003eFigure 8: Beeswarm plot to identify determinants of risky behavior among pregnant women 15–59 years in Eastern African countries, 2012-2022.\u003c/p\u003e","description":"","filename":"8.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7364399/v1/bd0afe0b2a68a9343be3924a.jpg"},{"id":91309644,"identity":"ffb12854-85a7-4b2e-8d7e-8c37463d7be3","added_by":"auto","created_at":"2025-09-15 07:07:51","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1716917,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7364399/v1/1cd7ad70-a790-4913-b372-d79e500fe6e1.pdf"},{"id":91308251,"identity":"09740db3-f69c-4b4c-b325-51951a2ca363","added_by":"auto","created_at":"2025-09-15 06:51:50","extension":"tif","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":140394,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eS1. Figure\u003c/strong\u003e. Map of study area Eastern African Countries to identify the determinants of risky behavior among pregnant women, 2012-2022.\u003c/p\u003e","description":"","filename":"S1.tif","url":"https://assets-eu.researchsquare.com/files/rs-7364399/v1/cb18090fe4e1370bc3c5f450.tif"},{"id":91307971,"identity":"2e74f394-965f-4a6b-b5be-c2ad4e4b070d","added_by":"auto","created_at":"2025-09-15 06:43:50","extension":"tif","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":74712,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eS2. Figure\u003c/strong\u003e. Sample size selection flow chart to identify the determinants of risky behavior among pregnant women, 2012-2022.\u003c/p\u003e","description":"","filename":"S2.tif","url":"https://assets-eu.researchsquare.com/files/rs-7364399/v1/88ca692a3eac24d0d3f0bb2f.tif"},{"id":91307972,"identity":"66c0c722-198a-4ee4-ad8b-ed674fc43a54","added_by":"auto","created_at":"2025-09-15 06:43:50","extension":"tif","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":166752,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eS5. Figure\u003c/strong\u003e. ROC curve analysis of selected machine learning algorithms to identify determinants of risky behavior among pregnant women 15–59 yearsin Eastern African countries, 2012-2022.\u003c/p\u003e","description":"","filename":"S5.tif","url":"https://assets-eu.researchsquare.com/files/rs-7364399/v1/761fdb5cb040c5d9f70f11cd.tif"},{"id":91307969,"identity":"3580686a-498d-49a0-bd71-a7cb411f0c89","added_by":"auto","created_at":"2025-09-15 06:43:50","extension":"pdf","order_by":4,"title":"","display":"","copyAsset":false,"role":"supplement","size":35465,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eS6.\u003c/strong\u003e \u003cstrong\u003ePDF.\u003c/strong\u003e DHS Permission Letter\u003c/p\u003e","description":"","filename":"S6.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7364399/v1/5bd3c645936591535dcb594a.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Exploring Machine learning Algorithms to Identify Determinants of Risky Behavior among Pregnant Women 15–59 years in Eastern African countries using the Demographic and Health Survey data","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eHealth-related risky behavior encompasses actions or personal traits that destroy health and well-being both in the short and long term. These behaviors, which include activities such as smoking, excessive alcohol use, engaging in harmful practices, physical inactivity, and engaging in risky sexual activity, are destructive health behaviors that may lead to poor health care (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eAt a global level, the majority of the burden of disease shifts to non-communicable diseases (NCDs) (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e, \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e). There are about 185\u0026nbsp;million drug users, 1.3\u0026nbsp;billion smokers, and 2\u0026nbsp;billion alcohol drinkers from those 9.8% of pregnant women consume alcohol, with higher prevalence in Europe (25.2%) and certain low- and middle-income countries (LMICs)(\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e) approximately 5.7% of pregnant women use tobacco(\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e). Additionally, 5.4% and 3.7% of the global burden of disease are related to alcohol and tobacco use respectively (\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e). The prevalence of substance use in a combined way was 80% globally (\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e). Similarly, Physical inactivity during pregnancy affects 28% of women(\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e). Collectively 7.2% of all deaths and 7.6% of cardiovascular deaths are linked to physical inactivity (\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e). In SSA, chewing khat among pregnant women ranges from 15\u0026ndash;30%(\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e). A study showed that substance use in Sub-Saharan Africa accounts to 41.6% and in Central Africa (55.5%) from those the most commonly used substances are caffeine products (41.2%), alcohol (32.8%), tobacco (23.5%), and khat (22%) (\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e). In Eastern African countries, substance use accounts for 43.70% (\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e), and khat users in African countries account for 22% (\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e), Comoros (23.90%), Uganda (65.60%), Ethiopia (68.70%), and Mozambique (76.70%) (\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e). Additionally, behaviors such as smoking, excessive alcohol consumption, and physical inactivity contribute significantly to the burden of disease and morbidity (\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e, \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eRisk behaviors have a great impact on women's mental, psychological, social, and physical health. The effect of this high prevalence of health risks leads to high premature death related to non-communicable diseases (NCDs) such as cardiovascular diseases, diabetes, cancer, high blood pressure, and respiratory problems associated with harmful alcohol consumption, poor dietary intake, cigarette use, and physical inactivity (\u003cspan additionalcitationids=\"CR18\" citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e). Similarly, risky behaviors contribute significantly to both communicable and non-communicable diseases (\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e, \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e, \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e). Moreover, risky behaviors among pregnant women have different health consequences, such as increased risks of chronic diseases, mortality, and disability, and a negative impact on physical and mental health (\u003cspan additionalcitationids=\"CR23\" citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eThe WHO and CDC are conducting a Global School Health Survey to identify behavioral risks and focus on promoting physical activity, healthy diets, and regulating unhealthy food marketing to children (\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e). Moreover, many interventions are conducted to get rid of risky behaviors. However, they targeted only a single behavior, while few have addressed multiple behaviors to reduce risky behaviors, although the effects are often small and may not be sustained long-term to reduce public health burdens (\u003cspan additionalcitationids=\"CR27 CR28 CR29 CR30 CR31\" citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e). Therefore, it is crucial to develop and evaluate appropriate interventions that target people with risky behaviors to help them refrain from engaging in risky behaviors. Existing efforts often focus on single issues, such as smoking or unhealthy diets, rather than tackling interrelated behaviors like physical inactivity, substance use, and poor nutrition simultaneously (\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e, \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e, \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eBased on an extensive literature search, no studies have been conducted on machine learning models to identify risky behavior among pregnant women 15\u0026ndash;59 years old. Additionally, no such studies have been conducted in Eastern African countries. Furthermore, most of the existing literature relied on traditional statistical models, limited geographic coverage, inability to assess the combined effects of multiple risk behaviors, and institution-based data records. Hence, to handle these issues, a machine learning approach could help to identify factors associated with risky behavior in a comprehensive way (\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eFurthermore, because risky behaviors are complex and multifaceted, traditional statistical models like logistic regression and linear regression usually fall short in capturing nonlinear patterns and interactions when used to support early health prevention strategies. (\u003cspan additionalcitationids=\"CR37\" citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e). Machine learning, on the other hand, can find new information and work with high-dimensional data. It can also give useful explanations through tools like SHAP(\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e).\u003c/p\u003e"},{"header":"2. Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003ch2\u003e2.1. Data source and study design\u003c/h2\u003e\u003cp\u003eThis study employed a population-based cross-sectional design utilizing secondary data extracted from the DHS conducted in 12 East African countries from 2012 to 2022. East Africa is a geographically diverse region extending from the Horn of Africa to parts of Southern Africa. This study used standard DHS data from 12 countries (Burundi (2016-17), Ethiopia (2016), Madagascar (2021), Comoros (2012), Rwanda (2019-20), Tanzania (2022), Mozambique (2022-23), Zimbabwe (2015), Kenya (2022), and Zambia (2015). The 12 nations were chosen based on the variables of interest being available in the respective databases (S1).\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec4\" class=\"Section2\"\u003e\u003ch2\u003e2.2. Population\u003c/h2\u003e\u003cp\u003eAll reproductive women 15\u0026ndash;59 years residing in 12 Eastern African countries were the source population. Whereas all pregnant women 15\u0026ndash;59 years living in the 12 Eastern African countries and present in the household during the enumeration period were the study population.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec5\" class=\"Section2\"\u003e\u003ch2\u003e2.3. Inclusion and exclusion criteria\u003c/h2\u003e\u003cp\u003eAll pregnant women aged 15\u0026ndash;59 years who had complete and valid responses for at least one of the risky behavior indicators in the DHS data in 12 Eastern African countries were included in the study. Whereas non-pregnant women aged 15\u0026ndash;59 years were excluded from the study.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec6\" class=\"Section2\"\u003e\u003ch2\u003e2.4 Sample size determination and sampling procedures\u003c/h2\u003e\u003cp\u003eA total of 199,083 reproductive women were included; of those, 184,746 women were not pregnant during the enumeration period. Finally, this study utilized a sample of 14,337 pregnant women from 12 Eastern African countries, who were selected according to the detailed flow chart for study participants, as presented in (S2).\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec7\" class=\"Section2\"\u003e\u003ch2\u003e2.5. Data collection\u003c/h2\u003e\u003cp\u003eDHS uses a standardized and validated questionnaire. It used a two-stage stratified sampling technique to select representative study participants. To begin with, the Enumeration Areas (EAs) were chosen using a probability method that was aligned with the size of each area, making sure the selection was done independently in every sampling group. In the next phase, homes were selected in a systematic way. The key demographic and health indicators were collected in each DHS (\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e). For this study, Individual Record (IR) datasets from the DHS from 2012 to 2022 were utilized. All the detailed information for the survey (such as the sampling method, the determination of the sample size, and the data collection procedure) is available in Demographic and Health Survey reports from the Measure DHS program website \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.dhsprogram.com\u003c/span\u003e\u003cspan address=\"https://www.dhsprogram.com\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\u003ch2\u003e2.6 Study variables\u003c/h2\u003e\u003cp\u003eThe outcome variable for the study was risky behavior.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e\u003ch2\u003e2.7. Operational Definition\u003c/h2\u003e\u003cp\u003eRisky behavior was dichotomized into two categories by merging two variables together. The women were considered to be engaging in risky behavior if they exhibited at least one of the following behaviors: using tobacco products (smoke cigarette, smoke pipe full of tobacco, snuffs by noise, snuffs by mouth, smoke kreteks, smoke cigar or cheroots, smoke water pipe, chewing tobacco, chewing betel liquid with tobacco, and smoke other local substance), and engaging in unsafe sexual activity (condom used during last sex from partners, number of sexual partner excluding spouse, ever been forced to unwanted sexual activity by partner). Whereas, the women were considered not engaged in risk behaviors if they exhibited none of the above behaviors (\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e, \u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e).\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec10\" class=\"Section2\"\u003e\u003ch2\u003e2.8. Data quality management\u003c/h2\u003e\u003cp\u003eThis study employed on utilizing high-quality data to which guarantee the validity and reliability of the model predictive performance. Some of the important parts of data preprocessing that were talked about were completeness, accuracy, uniqueness, timeliness, and consistency. For example, Stata made sure that the data was complete by filling in missing values before any more analysis. To improve the accuracy of the data and make sure the results were reliable, outliers in the dataset were found and removed using the right statistical methods.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\u003ch2\u003e2.9. Data management and analysis\u003c/h2\u003e\u003cp\u003eThis study used a machine learning approach based on Yufeng Guo\u0026rsquo;s 7 steps of ML and the frameworks of a previous study. The seven steps employed in the management and analysis include data collection, data preprocessing, model selection, model training, model evaluation, hyperparameter tuning, and making predictions (\u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e). Python version 3.7 on Jupyter Notebook and R were used for data preprocessing, modeling, and statistical analysis.\u003c/p\u003e\u003cp\u003e\u003cb\u003eData preprocessing\u003c/b\u003e\u003c/p\u003e\u003cp\u003eThe process of machine learning begins with data pre-processing, which involves modifying or encoding the data to make it suitable for computer interpretation (84). In machine learning workflow, this study was employed a continuous improvement process for models. This process included selecting and engineering relevant features, splitting the data, model training, and model evaluation, model optimization, choosing the top performer model, and deploying the selected model for prediction. Through an iterative approach, study was refining our models.\u003c/p\u003e\u003cp\u003e\u003cb\u003eData cleaning\u003c/b\u003e\u003c/p\u003e\u003cp\u003eMissing value in the dataset was inspected amd imputed by using the K-nearest neighbors (KNN) imputation was used to fill in the missing values. The KNN approach was specifically used to impute variables with less than 3% missing data, taking into account both mean and median values (\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e). Variables with more than 3% missing data were excluded from the analysis.Furthermore, multicollinearity was evaluated by using the correlation coefficient matrix; high multicollinearity was defined as having a correlation coefficient more than 0.8, hence the independent variables were not interfere with each other(\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e).\u003c/p\u003e\u003cp\u003e\u003cb\u003eDimensionality reduction\u003c/b\u003e\u003c/p\u003e\u003cp\u003eBased on shadow features importance for comparison, Boruta iteratively confirms or disproves attributes based on statistical evidence, ensuring that only significant predictors are retained. The Boruta algorithm was employed in this study as a dimensionality reduction technique to enhance model performance and reduce data complexity. High-dimensional datasets often contain repetitive or superfluous features, which can lower model accuracy and increase computational load. The Boruta algorithm was performed on R v4.33, then the feature importance was visualized using Boruta feature importance using boxplots. Non-important variables were rejected by the algorithm; and dimensionality reduction involves decreasing the number of input features to increased model efficiency, improve model performance(\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e).\u003c/p\u003e\u003cp\u003e\u003cb\u003eFeature engineering\u003c/b\u003e\u003c/p\u003e\u003cp\u003eFeature engineering including the process of identifying, categorizing, acquiring, and encoding the most relevant characteristics from the available datasets to build machine learning models that are more accurate (\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e). This study employed one-hot encoding for nominal categorical variables and label encoding for ordinal categorical variables to encode the data (\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e).\u003c/p\u003e\u003cp\u003e\u003cb\u003eData balancing\u003c/b\u003e\u003c/p\u003e\u003cp\u003eTo address the issue of data balancing for the outcome variable, the SMOTEENN (Synthetic Minority Over-Sampling Technique with the Edited Nearest Neighbor\u003cb\u003e)\u003c/b\u003e method was used. The SMOTEENN first generates additional synthetic samples for the minority class by including the existing samples within the feature space which helps to reducing the risk of overfitting.\u003c/p\u003e\u003cp\u003e\u003cb\u003eModel selection\u003c/b\u003e\u003c/p\u003e\u003cp\u003eIn our study, the dependent variable included a method for classifying risk behaviors. This study the appropriate classifier was selected based on the literatures that use before for the purpose of classification to make predictions of health risky behaviors. The study used the Scikit-learn package version 3.7 in Python running within Jupyter Notebook to implement the machine learning algorithm. The selection of these algorithms was based on their suitability for classification tasks and their compatibility with the characteristics of our dataset include Random Forest (RF), Logistic Regression (LR), LightGBM, Na\u0026iuml;ve bays, Artficial Neural Network (ANN) Decision Tree (DT), extreme gradient boosting (XGB), K-nearest neighbor (KNN), AdaBoost, support vector machine (SVM), Cat Boost was used in this study and implemented by python version 3.7 using each packages and on Jupiter notebook (\u003cspan additionalcitationids=\"CR50\" citationid=\"CR49\" class=\"CitationRef\"\u003e49\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e51\u003c/span\u003e).\u003c/p\u003e\u003cp\u003e\u003cb\u003eModel building and evaluation\u003c/b\u003e\u003c/p\u003e\u003cp\u003eTo make good predictive models, you need to train the model. The model learned from the training and testing data split of 30/70%, 20/80%, and 90/10%. This different way of splitting the data lets us accurately measure how well each prediction model works. The 80/20% train test data worked best for our model. Several metrics including sensitivity, specificity, and AUC were used to evaluate the performance of the prediction models. For this study accuracy and AUC was implemented to measure to evaluate effective model evaluation with imbalanced data (\u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e52\u003c/span\u003e). Each of these metrics provides valuable understanding into various parts of model performance.\u003c/p\u003e\u003cp\u003eAccuracy measures the overall correctness of the model\u0026rsquo;s predictions. It is the ratio of the number of correct predictions to total predictions. Precision quantifies the accuracy of positive predictions made by the model. It finds the ratio of true positive predictions to total predicted positives. Recall, also known as sensitivity or true positive rate, assesses the model\u0026rsquo;s ability to identify all positive instances it is effectively utilized in various computational models and machine learning applications (\u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e53\u003c/span\u003e). AUC is alternative measure other than accuracy to evaluating the performance of models, specifically in an unbalanced data sets, since it is threshold-independent and ability to provide a more overall performances for predictive models (\u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e54\u003c/span\u003e).\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eEvaluation metrics to identify the determinants of risky behavior among pregnant women, 2012\u0026ndash;2022.\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=\"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\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eClasses\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colspan=\"3\" nameend=\"c4\" namest=\"c2\"\u003e\u003cp\u003ePredicted 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\u003cp\u003eYES\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eNO\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eTotal\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eYes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ea (TN)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eb (FP)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eTN\u0026thinsp;+\u0026thinsp;FN\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNo\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ec (FN)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003ed (TP)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eFN\u0026thinsp;+\u0026thinsp;TP\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTotal\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTN\u0026thinsp;+\u0026thinsp;FN\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eFP\u0026thinsp;+\u0026thinsp;TP\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eTN\u0026thinsp;+\u0026thinsp;FN\u0026thinsp;+\u0026thinsp;FP\u0026thinsp;+\u0026thinsp;TP\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\u003eAs illustrated in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e two by two tables, the following formula were used to calculate.\u003c/p\u003e\u003cp\u003eAccuracy\u0026thinsp;=\u0026thinsp;TP\u0026thinsp;+\u0026thinsp;TN /TP\u0026thinsp;+\u0026thinsp;TN\u0026thinsp;+\u0026thinsp;FP\u0026thinsp;+\u0026thinsp;FN\u003c/p\u003e\u003cp\u003eSensitivity\u0026thinsp;=\u0026thinsp;TP /TP\u0026thinsp;+\u0026thinsp;FN\u003c/p\u003e\u003cp\u003eSpecificity\u0026thinsp;=\u0026thinsp;TN /TN\u0026thinsp;+\u0026thinsp;FP\u003c/p\u003e\u003cp\u003ePrecision\u0026thinsp;=\u0026thinsp;TP/ TP\u0026thinsp;+\u0026thinsp;FP\u003c/p\u003e\u003c/p\u003e\u003cp\u003eUtilizing the above metrics, the study comprehensively evaluated the performance of each predictive model in terms of overall correctness, accurate positive predictions, and identification of positive instances, balanced measure, and discriminatory ability.\u003c/p\u003e\u003cp\u003e\u003cb\u003eHyperparameter tuning\u003c/b\u003e\u003c/p\u003e\u003cp\u003eHyperparameter is an external manipulation to the model whose value must be set by the user (\u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e55\u003c/span\u003e). The selected model was optimized with the best parameters by applying the Optuna with 10-fold cross validation on the specified search space with one hundred trials. Since these techniques is more efficient when we deal with complex models and larger dataset than other techniques by using python version 3.7 on Jupiter notebook.\u003c/p\u003e\u003cp\u003e\u003cb\u003eModel interpretability\u003c/b\u003e\u003c/p\u003e\u003cp\u003eIn this study, the interpretability of machine learning models was essential to ensure clinical relevance and support evidence-based decision-making. To address the complexity and black-box nature of advanced predictive models, SHapley Additive exPlanations (SHAP) were employed. SHAP, based on cooperative game theory, offers a unified and theoretically grounded method to explain individual predictions by fairly attributing the contribution of each feature to the model\u0026rsquo;s output (\u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e56\u003c/span\u003e). Among the various SHAP implementations, Tree SHAP was selected for its computational efficiency and exact calculations when applied to tree-based models. This was especially relevant since Light GBM was identified as the best-performing model in this study.\u003c/p\u003e\u003cp\u003eSHAP played a central role in both feature selection and model interpretation. Features were levelled by considering their average absolute SHAP values, which handle to identify and the variability of the outcome variables. Moreover, a range of SHAP visualization techniques were used to improve both global and local model interpretability. The global feature importance plot provided an overall ranking of feature. Similarly, the beeswarm plot illustrated the direction and distribution of features across the population, and the waterfall plot offered detailed case-level explanations. These tools collectively facilitated the discovery of complex interactions and non-linear relationships between the predictors that are often overlooked using traditional statistical methods. In this study, the integration of SHAP provided a comprehensive analytical framework for understanding health risk behaviors among adult men in East Africa. SHAP enabled transparent and data-driven explanations of model predictions. The overall data preprocessing and analysis workflow is summarized and presented in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e3\u003c/span\u003e.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e"},{"header":"3. Result","content":"\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e\u003ch2\u003e3.1. Socio-demographic and Behavioral related characteristics of study participants\u003c/h2\u003e\u003cp\u003eMajority of the respondents have attained primary (47.7%), while 21.7% have no formal education. About 36.7% were categorized as rich, and 44.5% as poor. The majority reside in urban areas (74.8%). More than half of the respondents (54.5%) were currently working.\u003c/p\u003e\u003cp\u003eAbout 16.56% of the study participant were engaged in some form of risky behaviors. Approximately 14.76% of participants were involved in unsafe sexual activity, while 2.16% of participants were using tobacco products. All the detailed descriptions of study participants are presented in Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eSocio-demographic and Behavioral related characteristics of pregnant women in Eastern African countries, 2012\u0026ndash;2022 (n\u0026thinsp;=\u0026thinsp;14337).\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\u003eVariable\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eCategories\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eFrequency\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colspan=\"2\" nameend=\"c5\" namest=\"c4\"\u003e\u003cp\u003ePercentage\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"2\" rowspan=\"3\"\u003e\u003cp\u003eWomen\u0026rsquo;s age\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e15\u0026ndash;24\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e5486\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c5\" namest=\"c4\"\u003e\u003cp\u003e38.21\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e25\u0026ndash;36\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e7550\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c5\" namest=\"c4\"\u003e\u003cp\u003e32.7\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e\u0026gt;=37\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1300\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c5\" namest=\"c4\"\u003e\u003cp\u003e29.1\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"2\" rowspan=\"3\"\u003e\u003cp\u003ewomen\u0026rsquo;s education\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eNo education\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e2958\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c5\" namest=\"c4\"\u003e\u003cp\u003e20.6\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ePrimary\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e6846\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c5\" namest=\"c4\"\u003e\u003cp\u003e47.7\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eSecondary \u0026amp; Higher\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e4535\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c5\" namest=\"c4\"\u003e\u003cp\u003e31.6\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"2\" rowspan=\"3\"\u003e\u003cp\u003eMarital Status\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eSingle\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1061\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c5\" namest=\"c4\"\u003e\u003cp\u003e7.4\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eDivorced/separated\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e712\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c5\" namest=\"c4\"\u003e\u003cp\u003e87.6\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eMarried\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e12564\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c5\" namest=\"c4\"\u003e\u003cp\u003e4.9\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eWomen\u0026rsquo;s Occupation\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eNot-working\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e6522\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c5\" namest=\"c4\"\u003e\u003cp\u003e45.9\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eWorking\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e7815\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c5\" namest=\"c4\"\u003e\u003cp\u003e54.5\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"2\" rowspan=\"3\"\u003e\u003cp\u003eHusband\u0026rsquo;s Education\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eNo-Education\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e2211\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c5\" namest=\"c4\"\u003e\u003cp\u003e17.6\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ePrimary\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e5441\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c5\" namest=\"c4\"\u003e\u003cp\u003e43.34\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eSecondary \u0026amp; Higher\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e4830\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c5\" namest=\"c4\"\u003e\u003cp\u003e40\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eANC Visits\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eNo\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e7190\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c5\" namest=\"c4\"\u003e\u003cp\u003e51.93\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eYes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e7146\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c5\" namest=\"c4\"\u003e\u003cp\u003e48\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"2\" rowspan=\"3\"\u003e\u003cp\u003eWealth status\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ePoor\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e6383\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c5\" namest=\"c4\"\u003e\u003cp\u003e44.5\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eMiddle\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e2688\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c5\" namest=\"c4\"\u003e\u003cp\u003e18.7\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eRich\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e5266\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c5\" namest=\"c4\"\u003e\u003cp\u003e36.7\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eHousehold head sex\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eMale\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e5699\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c5\" namest=\"c4\"\u003e\u003cp\u003e74.97\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eFemale\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1902\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c5\" namest=\"c4\"\u003e\u003cp\u003e25.03\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"2\" rowspan=\"3\"\u003e\u003cp\u003eFamily size\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1_4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e7022\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c5\" namest=\"c4\"\u003e\u003cp\u003e48.9\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e5\u0026ndash; 8\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e5899\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c5\" namest=\"c4\"\u003e\u003cp\u003e41.1\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e\u0026gt;=9\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1416\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c5\" namest=\"c4\"\u003e\u003cp\u003e9.88\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eFuel type\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eNot-Electricity\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e13635\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c5\" namest=\"c4\"\u003e\u003cp\u003e95.1\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eElectricity\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e702\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c5\" namest=\"c4\"\u003e\u003cp\u003e4.9\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eMedia exposure\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eNo\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e5427\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c5\" namest=\"c4\"\u003e\u003cp\u003e37.8\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eYes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e8910\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c5\" namest=\"c4\"\u003e\u003cp\u003e62.1\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003ePlace of residence\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eUrban\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e10736\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e74.88\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eRural\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e3601\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e25.12\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eRisky behavior\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eYes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e2374\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c5\" namest=\"c4\"\u003e\u003cp\u003e16.56\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eNo\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e11963\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c5\" namest=\"c4\"\u003e\u003cp\u003e83.44\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eUnsafe sexual activity\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eYes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e2115\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c5\" namest=\"c4\"\u003e\u003cp\u003e14.75\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eNo\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e12222\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c5\" namest=\"c4\"\u003e\u003cp\u003e85.25\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eTobacco use\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eYes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e309\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c5\" namest=\"c4\"\u003e\u003cp\u003e2.16\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eNo\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e14028\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c5\" namest=\"c4\"\u003e\u003cp\u003e97.84\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e\u003ch2\u003e3.2 Machine learning analysis results\u003c/h2\u003e\u003cdiv id=\"Sec15\" class=\"Section3\"\u003e\u003ch2\u003e3.2.1 Feature selection\u003c/h2\u003e\u003cp\u003eIn this study, Boruta analysis identified several significant predictors of risky behavior. Features such as country of residence, age, sex of household head, mobile phone use, residence, marital status, and media exposure were consistently confirmed as important, as indicated by their high importance scores in the final boxplot output. These variables demonstrated a strong contribution to the model's predictive capacity. This guided the final modeling phase to focus on features with confirmed relevance, reducing noise and improving overall model performance. The figure below presents the final importance scores of the Boruta algorithm. Based on the Boruta algorithm, features that were represented with the green color and presented right to the Shadomax were included in the next analysis. However, marital status and family size were unimportant for model building and rejected by the algorithm as presented in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e4\u003c/span\u003e.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec16\" class=\"Section3\"\u003e\u003ch2\u003e3.2.2 Data balancing\u003c/h2\u003e\u003cp\u003eIn this study, the SMOTENN (Synthetic Minority Over-Sampling Technique combined with Edited Nearest Neighbors) algorithm was used to solve the problem of class imbalance in the outcome variable. The SMOTENN oversampling technique generated 9589 additional synthetic data for minority classes hence the data was changed from unbalanced to balanced distribution for both classes. The updated proportions greatly decreased the discrepancy while maintaining data integrity, even though the outcome did not produce a perfect 50:50 balance. This result was obtained by combining artificial oversampling of the minority class with ENN's elimination of noisy or ambiguous examples from the majority class. Consequently, the balanced dataset made it easier to train a more robust and equitable model, which improved performance as shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e5\u003c/span\u003e.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec17\" class=\"Section3\"\u003e\u003ch2\u003e3.2.3 Model development and performance evaluation to predict risky behavior\u003c/h2\u003e\u003cp\u003eBased on the comprehensive evaluation of model performance metrics (Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e), Light GBM demonstrates the highest performance among all models, achieving the top accuracy of 90.6% and the highest ROC AUC of 96.8%. XGBoost and Neural Network also performed competitively, with accuracies of 89.6% and 88.4%, and ROC AUC values of 0.9656 and 0.9497as illustrated in ROC curve results in (S5), respectively. In contrast, traditional models such as Logistic Regression and Naive Bayes showed relatively weaker performance, with accuracies below 70% and lower precision.\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\u003eEvaluation matrices of balanced to identify determinants of risky behavior among pregnant women 15\u0026ndash;59 years in Eastern African countries, 2012\u0026ndash;2022.\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"6\"\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\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\u003eML model\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eAccuracy score (%)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003ePrecision\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eSensitivity\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c6\"\u003e\u003cp\u003eAUC score\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e1.\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eSupport Vector Machine (SVM)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e88\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.8579\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.76425\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.7657\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e2.\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e\u003cb\u003eLightGBM\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e\u003cb\u003e90\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cb\u003e0.9018\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e\u003cb\u003e0.8666\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e\u003cb\u003e0.9680*\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e3.\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eAdaBoost\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e89\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.8754\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.9301\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.8545\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e4.\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eCat boost\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e88.9\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.8984\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.8424\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.7545\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e5.\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eXGBoost\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e89.5\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.8912\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.8515\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.9656\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e6.\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eNeural Network (ANN)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e88.4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.8698\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.8455\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.9497\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e7.\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eKNN\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e84.6\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.8349\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.738\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.9342\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e8.\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eRandom Forest (RF)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e74.6\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.7855\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.5322\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.8347\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e9.\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eDecision Tree (DT)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e71.8\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.6465\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.699\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.8031\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e10.\u0026nbsp;\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eNaive Bayes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e68\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.6417\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.5119\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.7248\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e11.\u0026nbsp;\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eLogistic Regression (LR)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e66.2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.6272\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.4494\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.7067\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e\u003cp\u003eMaximum model performance *\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec18\" class=\"Section3\"\u003e\u003ch2\u003e3.2.4 Hyperparameter tuning for the best model\u003c/h2\u003e\u003cp\u003eAfter evaluating the performance of the model, Light GBM was the best-performing model in terms of both predictive accuracy and area under the curve. To further optimize Light GBM\u0026rsquo;s performance, we applied Bayesian Optimization using the Optuna framework after optimizing the default and optimized values are provided in Table\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e.\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\u003eDefault and optimal hyperparameters for Light GBM model to identify determinants of risky behavior among pregnant women 15\u0026ndash;59 years in Eastern African countries, 2012\u0026ndash;2022.\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=\"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\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\u003eHyperparameters\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eDefault value\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eOptimum value\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\u003eboosting_type\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e\u0026lsquo;gdbt\u0026rsquo;\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u0026lsquo;dart\u0026rsquo;\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\u003eObjectives\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eregression\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eBinary\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\u003eMetrics\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e12\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eAuc\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\u003en_estimators\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e100\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e300\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\u003enum_leaves\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e31\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e100\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\u003elearning_rates\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.1\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.3\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\u003emax_depth\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e-1\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e7\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec19\" class=\"Section3\"\u003e\u003ch2\u003e3.2.5 SHAP Summary Plot result\u003c/h2\u003e\u003cp\u003eA SHAP (SHapley Additive exPlanations) summary bar plot was created to comprehend the overall contribution of each feature in predicting risk behavior (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e7\u003c/span\u003e). The mean absolute SHAP value for each of the top 9 most significant variables is displayed in this plot, indicating the average degree of influence each feature has on the model output for every individual, whether in a positive or negative direction. The mean SHAP value is shown on the x-axis, and the associated features are listed on the y-axis. Variable with more mean absolute SHAP values or variables appear on top of the bar reveal the most important predictors such as: women who lived in rural area (place residence_0), women\u0026rsquo;s with poor wealth income (wealth status_0), women\u0026rsquo;s with middle wealth income(wealth status_1), husbands education being primary(husbands education_1), women\u0026rsquo;s being employed(women\u0026rsquo;s occupation _1), women\u0026rsquo;s being not exposed to media (media_exposure_0), women\u0026rsquo;s being utilize ANC services (ANC visit_1), women\u0026rsquo;s age being 25\u0026ndash;36(women\u0026rsquo;s age_1) are the top most important variables sorted in descending order from higher to lower as illustrated in Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e7\u003c/span\u003e.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec20\" class=\"Section3\"\u003e\u003ch2\u003e3.2.6 SHAP Beeswarm Plot result\u003c/h2\u003e\u003cp\u003eA SHAP (SHapley Additive exPlanations) beeswarm plot was used to provide a rich overview of how the variables impact the model\u0026rsquo;s predictions across all the data. The color denotes the feature value's magnitude (blue for low, red for high), and the x-axis represents the SHAP value, which quantifies the influence of a feature on the model output.\u003c/p\u003e\u003cp\u003eThe analysis was identified women who lived in rural area (place residence_0), women\u0026rsquo;s with poor wealth income (wealth status_0), women\u0026rsquo;s with middle wealth income (wealth status_1), husband\u0026rsquo;s education being primary (husbands education_1), and women\u0026rsquo;s being not exposed to media (media_exposure_0) increase the likelihood of risky behavior among pregnant women. Whereas women\u0026rsquo;s being employed (women\u0026rsquo;s occupation _1), women\u0026rsquo;s being utilized ANC services (ANC visit_1), and women\u0026rsquo;s age being 25\u0026ndash;36 (women\u0026rsquo;s age_1) lowers the likelihood of risky behaviors among pregnant women, as illustrated in Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e8\u003c/span\u003e.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e"},{"header":"4. Discussion","content":"\u003cp\u003eThis study aimed to identify determinants of risky behaviors among pregnant women in Eastern African countries using various machine learning models applied on the DHS data. For this purpose, 11 machine learning models were trained on the balanced training data through tenfold cross-validation. The performance of those 11 classifier models was compared by their classification accuracy and AUC score. During the first phase of predictive modeling on balanced training data, Light GBM performed better than other classifiers with an accuracy of 90.6% and a 0.968 AUC score. Hence, light GBM was the best predictive model and further analysis was performed after optimizing it for its optimal parameters. Light GBM demonstrated a strong ability to distinguish between individuals with and without risky behaviors. Similarly, a study conducted in China suggested that gradient boosting is the best model to predict risky sexual behaviors among university students (\u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e57\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eBased on the study finding that women living in rural areas had the highest probability of engaging in risky behaviors compared to urban. This finding is consistent with a study conducted in Virginia(\u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e58\u003c/span\u003e), Myanmar (\u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e59\u003c/span\u003e) and Ethiopia (\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e). This might be due to the traditional lifestyles, lack exposed to health education, inadequate access nutrition, or insufficient prenatal care less regulated access to drugs in rural areas, cultural norms that promote alcohol consumption, and a lack of public health initiatives could be contributing to the rise in risky behavior (\u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e60\u003c/span\u003e, \u003cspan citationid=\"CR61\" class=\"CitationRef\"\u003e61\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eWomen being from lower and middle wealth indices were more likely to engage in risky behaviors. This study is consistent with the study conducted in Canada (\u003cspan citationid=\"CR62\" class=\"CitationRef\"\u003e62\u003c/span\u003e), Denmark (\u003cspan citationid=\"CR63\" class=\"CitationRef\"\u003e63\u003c/span\u003e), and Pomerania (\u003cspan citationid=\"CR64\" class=\"CitationRef\"\u003e64\u003c/span\u003e). This could be because financial constraints may limit access to health-related information, wholesome food, and first-rate medical care. Women may also get stressed and frustrated due to financial hardship. Additionally, women with lower wealth levels may be more vulnerable to factors like limited access to services, financial stress, or living in lower socioeconomic status conditions (\u003cspan citationid=\"CR65\" class=\"CitationRef\"\u003e65\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eMoreover, paternal education being primary increases the probability of risky behaviors among pregnant women. This finding aligns with other studies conducted in Uganda (\u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e66\u003c/span\u003e) and Australia (\u003cspan citationid=\"CR67\" class=\"CitationRef\"\u003e67\u003c/span\u003e), showing that paternal education level significantly affects maternal health conditions. This might be due to male partner education being lower might be associated with the delay in intention to detect early pregnancy signs and symptoms, lower health care decision making, decreased support and counseling for pregnancy risks, and insufficient support for maternal healthful behavior (\u003cspan citationid=\"CR68\" class=\"CitationRef\"\u003e68\u003c/span\u003e, \u003cspan citationid=\"CR69\" class=\"CitationRef\"\u003e69\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eFurthermore, women who were not exposed to the media were more likely to engage in risky activities (\u003cspan citationid=\"CR70\" class=\"CitationRef\"\u003e70\u003c/span\u003e). This may be because the media is an important medium to transmitting essential public health messages, such as adequate dietary intake for pregnant mothers, ANC service utilization, environmental hygiene and sanitation, raising awareness among women on early pregnancy complications, and strengthening ANC care practices (\u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e71\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eWhereas women who are employed or working women decrease risky behaviors during pregnancy. This study was conducted in Australia (\u003cspan citationid=\"CR67\" class=\"CitationRef\"\u003e67\u003c/span\u003e) and Iran (\u003cspan citationid=\"CR72\" class=\"CitationRef\"\u003e72\u003c/span\u003e). This may be due to occupation can strengthen women's finances and increase their reliance on making a healthy lifestyle, reducing stress, including utilization of ANC care services on a timely basis, and promoting better healthful behaviors (\u003cspan citationid=\"CR67\" class=\"CitationRef\"\u003e67\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eSimilarly, women who attend ANC care services were less likely to engage in risky behaviors. This finding is aligned with other studies conducted in Iran (\u003cspan citationid=\"CR73\" class=\"CitationRef\"\u003e73\u003c/span\u003e) and Pomerania (\u003cspan citationid=\"CR64\" class=\"CitationRef\"\u003e64\u003c/span\u003e). This could be due to that ANC visits providing opportunities for health education, early detection of pregnancy complications, empowering women\u0026rsquo;s to have adequate knowledge about proper nutrition, substance use and provide counseling service on nutrition and substance avoidance, and empowering women\u0026rsquo;s to make better health decision all of which contribute to reducing risky behaviors (\u003cspan citationid=\"CR74\" class=\"CitationRef\"\u003e74\u003c/span\u003e, \u003cspan citationid=\"CR75\" class=\"CitationRef\"\u003e75\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eAdditionally, older women aged 25 and 36 were less likely to engage in risky behaviors. This age group is strongly associated to higher maturity and self-regulation as well as they have strong background for unhealthy patterns and better ability for impulse control and unwanted behaviors than younger women in handling, and physiological well-being for pregnancy. This finding is supported by studies conducted in Denmark (\u003cspan citationid=\"CR63\" class=\"CitationRef\"\u003e63\u003c/span\u003e), Ethiopia (\u003cspan citationid=\"CR76\" class=\"CitationRef\"\u003e76\u003c/span\u003e), Scotland (\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e), and South Africa (\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e, \u003cspan citationid=\"CR77\" class=\"CitationRef\"\u003e77\u003c/span\u003e). This could be because younger women (\u0026lt;\u0026thinsp;25 years) may lack experience, while older women may engage in less risky behaviors. This might be due to older women have a lower probability of engaging in risky behaviors like unsafe sex, transactional sex, use of substance use, or peer pressure influence (\u003cspan citationid=\"CR78\" class=\"CitationRef\"\u003e78\u003c/span\u003e, \u003cspan citationid=\"CR79\" class=\"CitationRef\"\u003e79\u003c/span\u003e).\u003c/p\u003e"},{"header":"5. Strengths and Limitations","content":"\u003cp\u003eThis study employed standardized measurements for measuring risky behaviors, which is internationally recognized and adopted by the WHO and other organizations widely used as a standardized tool for measuring risky behaviors in this study. Consequently, comparison of result across many studies will be easy.\u003c/p\u003e\u003cp\u003eSimilarly, the study used SHAP model interpretation techniques which enables us to interpret complex models accurately, which bases on an optimized and best-performing model, which increases model performance to accurately identify risky behaviors among pregnant women.\u003c/p\u003e\u003cp\u003eHowever, the study design was cross-sectional it does not show a causal relationship between the variables (cause-and-effect), similarly, it also does not show long-term effects.\u003c/p\u003e\u003cp\u003eMoreover, the secondary nature of the dataset makes it difficult to explore and find additional essential features to accurately predict and identify associated factors for risky behavior in compressively.\u003c/p\u003e"},{"header":"6. Conclusion","content":"\u003cp\u003eThis study aims to identify determinants of risky behavior among pregnant women aged 15\u0026ndash;59 years in Eastern African countries. The light GBM notably showed an outstanding performance with an accurate score of 95.88% and an AUC score of 0.991. Similarly, the findings revealed that women who lived in rural areas, women with poor wealth income, women with middle wealth income, husbands' education being primary, and women not being exposed to media increase the likelihood of risky behavior among pregnant women. Whereas, women\u0026rsquo;s being employed, women utilized ANC services, women\u0026rsquo;s age being 25\u0026ndash;36 lowers the likelihood of risky behaviors among pregnant women.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cp\u003eANC \u0026ndash; Ante Natal Care\u003c/p\u003e\n\u003cp\u003eANN \u0026ndash; Artificial neural Networks\u003c/p\u003e\n\u003cp\u003eAUC - Area under the Curve\u003c/p\u003e\n\u003cp\u003eCV - Cross Validation\u003c/p\u003e\n\u003cp\u003eCDC \u0026ndash; Communicable Disease Control\u003c/p\u003e\n\u003cp\u003eDHS - Demographic and Health Survey\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eDT - Decision Tree\u003c/p\u003e\n\u003cp\u003eEA - East Africa\u003c/p\u003e\n\u003cp\u003eFN - False Negative\u003c/p\u003e\n\u003cp\u003eFP \u0026ndash; False Positive\u003c/p\u003e\n\u003cp\u003eGBM \u003cstrong\u003e-\u003c/strong\u003e Light Gradient Boosting Machine\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eIR \u0026ndash; Individual Record\u003c/p\u003e\n\u003cp\u003eKNN - K-nearest neighbor\u003c/p\u003e\n\u003cp\u003eLR - Logistic Regression\u003c/p\u003e\n\u003cp\u003eML - Machine Learning\u003c/p\u003e\n\u003cp\u003eNCD \u0026ndash; Non-communicable Disease\u003c/p\u003e\n\u003cp\u003eRF - Random Forest\u003c/p\u003e\n\u003cp\u003eROC - Receiving Operating Characteristics\u003c/p\u003e\n\u003cp\u003eSHAP - Shapely Additive Explanation\u003c/p\u003e\n\u003cp\u003eSSA - Sub-Saharan Africa\u003c/p\u003e\n\u003cp\u003eSVM - Support Vector Machine\u003c/p\u003e\n\u003cp\u003eSMOTEENN - Synthetic Minority Oversampling Techniques with the Edited Nearest Neighbor\u003c/p\u003e\n\u003cp\u003eTN - True Negative\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTP \u0026ndash; True Positive\u003c/p\u003e\n\u003cp\u003eUNICEF - United Nations International Children\u0026rsquo;s Emergency Fund\u003c/p\u003e\n\u003cp\u003eWHO - World Health Organization\u003c/p\u003e\n\u003cp\u003eXGB - Extreme Gradient Boosting\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAuthors\u0026rsquo; Contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eH.W.J., did the conceptualization and investigation, A.B.K., S.W.S., and H.W.J. did data curation, H.W.J. and S.W.S. wrote the methodology. H.W.J., carried out the analysis. SW.S. H.W.J. wrote the original draft, A.E.B., A.B.K., and S.W.S. wrote and edited the final manuscript text. All authors subsequently revised the final manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConflict of interest\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare that they have no conflict of interest.\u003c/p\u003e\n\u003cp id=\"_Toc199416421\"\u003e\u003cstrong\u003eAvailability of data and materials\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll relevant data are within the manuscript and its Supporting Information files.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for publication\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthical approval and consent from participants\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eData was obtained after granted permission through proper registration to access the data from the DHS website https://www.dhsprogram.com. The data for this study was downloaded after full authorization and approval, and consent from the DHS committee with a formal letter, which is attached at the end of the document.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNo funding was obtained for this study.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors acknowledge to Measure of DHS program committee for authorizing us to use the data sets.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003e\u0026Aring;sberg K, Lundgren O, Henriksson H, Henriksson P, Bendtsen P, L\u0026ouml;f M, et al. Multiple lifestyle behaviour mHealth intervention targeting Swedish college and university students: protocol for the Buddy randomised factorial trial. BMJ Open. 2021;11(12):e051044.\u003c/li\u003e\n\u003cli\u003eDuffy KA, Green PA, Chartrand TL. Mimicry and modeling of health (-risk) behaviors: How others impact our health (-risk) behaviors without our awareness. Journal of Nonverbal Behavior. 2020;44(1):5-40.\u003c/li\u003e\n\u003cli\u003eBourne RR. Global burden of 369 diseases and injuries in 204 countries and territories, 1990\u0026ndash;2019: a systematic analysis for the Global Burden of Disease Study 2019. 2023.\u003c/li\u003e\n\u003cli\u003eBhattacharya S, Heidler P, Varshney S. Incorporating neglected non-communicable diseases into the national health program\u0026mdash;A review. Frontiers in public health. 2023;10:1093170.\u003c/li\u003e\n\u003cli\u003eAlcohol WGSRo. Global status report on alcohol and health 2018: Alcohol, Drugs and Addictive Behaviors (ADA); 2018 [\u003c/li\u003e\n\u003cli\u003eCDC. CDC Morbidity and mortality weekley report2019.\u003c/li\u003e\n\u003cli\u003eDegenhardt L, Charlson F, Ferrari A, Santomauro D, Erskine H, Mantilla-Herrara A, et al. The global burden of disease attributable to alcohol and drug use in 195 countries and territories, 1990\u0026amp;#x2013;2016: a systematic analysis for the Global Burden of Disease Study 2016. The Lancet Psychiatry. 2018;5(12):987-1012.\u003c/li\u003e\n\u003cli\u003eGlantz MD, Bharat C, Degenhardt L, Sampson NA, Scott KM, Lim CC, et al. The epidemiology of alcohol use disorders cross-nationally: Findings from the World Mental Health Surveys. Addictive behaviors. 2020;102:106128.\u003c/li\u003e\n\u003cli\u003eOrganization WH. Global status report on alcohol and health 2018: World Health Organization; 2018.\u003c/li\u003e\n\u003cli\u003eWHO. WHO guidelines on physical activity and sedentary behaviour2020.\u003c/li\u003e\n\u003cli\u003eKatzmarzyk PT, Friedenreich C, Shiroma EJ, Lee IM. Physical inactivity and non-communicable disease burden in low-income, middle-income and high-income countries. Br J Sports Med. 2022;56(2):101-6.\u003c/li\u003e\n\u003cli\u003eSreeramareddy CT, Pradhan PM, Sin S. Prevalence, distribution, and social determinants of tobacco use in 30 sub-Saharan African countries. BMC medicine. 2014;12:1-13.\u003c/li\u003e\n\u003cli\u003eFentaw KD, Fenta SM, Biresaw HB. Prevalence and Associated factors of Substance Use Male Population in East African countries: a multilevel analysis of recent demographic and health surveys from 2015 to 2019. Substance abuse: research and treatment. 2022;16:11782218221101011.\u003c/li\u003e\n\u003cli\u003e\u0026lt;substance-use-among-adolescents-in-sub-saharan-africa-a-systematic-review-and-meta-analysis.pdf\u0026gt;.\u003c/li\u003e\n\u003cli\u003ePark JH, Moon JH, Kim HJ, Kong MH, Oh YH. Sedentary lifestyle: overview of updated evidence of potential health risks. Korean journal of family medicine. 2020;41(6):365.\u003c/li\u003e\n\u003cli\u003ePurba MAK, Thomson DRM, Henery DPM, Pearce DA, Henderson PM, Katikireddi PSV. Social media use and adolescent health-risk behaviours: A systematic review and meta-analysis. medRxiv. 2023:2023.02. 24.23286415.\u003c/li\u003e\n\u003cli\u003eBista B, Dhimal M, Bhattarai S, Neupane T, Xu YY, Pandey AR, et al. Prevalence of non-communicable diseases risk factors and their determinants: Results from STEPS survey 2019, Nepal. PloS one. 2021;16(7):e0253605.\u003c/li\u003e\n\u003cli\u003eOlajide D, Eberth B, Ludbrook A. Analysis of multiple health risky behaviours and associated disease outcomes using Scottish linked hospitalisation data. Frontiers in public health. 2022;10:847938.\u003c/li\u003e\n\u003cli\u003eMeader N, King K, Moe-Byrne T, Wright K, Graham H, Petticrew M, et al. A systematic review on the clustering and co-occurrence of multiple risk behaviours. BMC public health. 2016;16:1-9.\u003c/li\u003e\n\u003cli\u003eFast D, Bukusi D, Moyer E. The knife\u0026apos;s edge: masculinities and precarity in East Africa. Social Science \u0026amp; Medicine. 2020;258:113097.\u003c/li\u003e\n\u003cli\u003eHuijts T, Gkiouleka A, Reibling N, Thomson KH, Eikemo TA, Bambra C. Educational inequalities in risky health behaviours in 21 European countries: findings from the European social survey (2014) special module on the social determinants of health. The European Journal of Public Health. 2017;27(suppl_1):63-72.\u003c/li\u003e\n\u003cli\u003eOkereke O. INFLUENCE OF PERSONALITY ON SEXUAL RISK BEHAVIOUR. Journal of Institutional Research, Big Data Analytics and Innovation. 2025;1(2):17-24.\u003c/li\u003e\n\u003cli\u003eFernandez A, Lozano A, Lee TK, Estrada Y, Messiah SE, Prado G. A family-based healthy lifestyle intervention: Crossover effects on substance use and sexual risk behaviors. Prevention Science. 2021;22:602-8.\u003c/li\u003e\n\u003cli\u003eGardner LA, Champion KE, Parmenter B, Grummitt L, Chapman C, Sunderland M, et al. Clustering of six key risk behaviors for chronic disease among adolescent females. International Journal of Environmental Research and Public Health. 2020;17(19):7211.\u003c/li\u003e\n\u003cli\u003ewho. Health education and promotion-Behavioural risk factors 2024 [Available from: (https://www.emro.who.int/health-education/health-risk-factors/behaviour-risk-factors.html).\u003c/li\u003e\n\u003cli\u003eSilva CC, Presseau J, van Allen Z, Schenk PM, Moreto M, Dinsmore J, et al. Effectiveness of interventions for changing more than one behavior at a time to manage chronic conditions: A systematic review and meta-analysis. Annals of Behavioral Medicine. 2024;58(6):432-44.\u003c/li\u003e\n\u003cli\u003eBourke M, Patten RK, Dash S, Pascoe M, Craike M, Firth J, et al. The effect of interventions that target multiple modifiable health behaviors on symptoms of anxiety and depression in young people: a meta-analysis of randomized controlled trials. Journal of Adolescent Health. 2022;70(2):208-19.\u003c/li\u003e\n\u003cli\u003eTzelepis F, Mitchell A, Wilson L, Byrnes E, Haschek A, Leigh L, et al. The long-term effectiveness of internet-based interventions on multiple health risk behaviors: systematic review and robust variance estimation meta-analysis. Journal of Medical Internet Research. 2021;23(12):e23513.\u003c/li\u003e\n\u003cli\u003eTinner L, Palmer JC, Lloyd EC, Caldwell DM, MacArthur GJ, Dias K, et al. Individual-, family-and school-based interventions to prevent multiple risk behaviours relating to alcohol, tobacco and drug use in young people aged 8-25 years: a systematic review and meta-analysis. BMC Public Health. 2022;22(1):1111.\u003c/li\u003e\n\u003cli\u003eDaniels KJ, Hoosen I, Pharaoh H. Health risk behaviour prevention/intervention programmes targeted at youth/adolescents engaging in risky behaviour\u0026mdash;A scoping review. Journal of Public Health. 2024:1-10.\u003c/li\u003e\n\u003cli\u003eG\u0026oacute;mez-G\u0026oacute;mez I, Moreno-Peral P, L\u0026oacute;pez T, Claver\u0026iacute;a A, Oliv\u0026aacute;n B, Marti R, et al. Multiple risk behaviour intervention to prevent depression in primary care. British Journal of General Practice. 2020;70(suppl 1).\u003c/li\u003e\n\u003cli\u003eChampion KE, Parmenter B, McGowan C, Spring B, Wafford QE, Gardner LA, et al. Effectiveness of school-based eHealth interventions to prevent multiple lifestyle risk behaviours among adolescents: a systematic review and meta-analysis. The Lancet Digital Health. 2019;1(5):e206-e21.\u003c/li\u003e\n\u003cli\u003eBoyle MA, Hoffmann AN, Horn J, Badger S, Gaskill LA. Approaches for Treating Multiply Controlled Problem Behavior. Behavior Analysis in Practice. 2024;17(1):53-69.\u003c/li\u003e\n\u003cli\u003eNoh S, Choo J. Gender-Specific Clustering of Lifestyle Behaviors and Its Impacts on Cardiovascular Health. Journal of Cardiovascular Nursing. 2024;39(4):E115-E25.\u003c/li\u003e\n\u003cli\u003eUddin S, Khan A, Hossain ME, Moni MA. Comparing different supervised machine learning algorithms for disease prediction. BMC medical informatics and decision making. 2019;19(1):1-16.\u003c/li\u003e\n\u003cli\u003ePasha SJ, Mohamed ES. Novel Feature Reduction (NFR) model with machine learning and data mining algorithms for effective disease risk prediction. Ieee Access. 2020;8:184087-108.\u003c/li\u003e\n\u003cli\u003eNaveenkumar S, Kirubhakaran R, Jeeva G, Shobana M, Sangeetha K. Smart health prediction using machine learning. International Research Journal on Advanced Science Hub (IRJASH). 2021;3(3):124-8.\u003c/li\u003e\n\u003cli\u003eMorgenstern JD, Buajitti E, O\u0026rsquo;Neill M, Piggott T, Goel V, Fridman D, et al. Predicting population health with machine learning: a scoping review. BMJ open. 2020;10(10):e037860.\u003c/li\u003e\n\u003cli\u003eMateen BA, Liley J, Denniston AK, Holmes CC, Vollmer SJ. Improving the quality of machine learning in health applications and clinical research. Nature Machine Intelligence. 2020;2(10):554-6.\u003c/li\u003e\n\u003cli\u003eCroft TN, Marshall AM, Allen CK, Arnold F, Assaf S, Balian S. Guide to DHS statistics. Rockville: ICF. 2018;645:292-303.\u003c/li\u003e\n\u003cli\u003eKebede SA, Weldesenbet AB, Tusa BS. Spatial Distribution and Factors Associated With Risky Health Behavior Among Adult Males Aged 15\u0026ndash;59 Years in Ethiopia: Generalized Structural Equation Modeling. Frontiers in Psychiatry. 2022;13:688336.\u003c/li\u003e\n\u003cli\u003eBrener ND, Collins JL. Co-occurrence of health-risk behaviors among adolescents in the United States. J Adolesc Health. 1998;22(3):209-13.\u003c/li\u003e\n\u003cli\u003eKumar M, Yadav P, Singh H, Arora A, editors. Efficient Methods to Improve the Performance of Supervised Learning Models. 2021 International Conference on Intelligent Technologies (CONIT); 2021: IEEE.\u003c/li\u003e\n\u003cli\u003eRawat S, Rawat A, Kumar D, Sabitha AS. Application of machine learning and data visualization techniques for decision support in the insurance sector. International Journal of Information Management Data Insights. 2021;1(2):100012.\u003c/li\u003e\n\u003cli\u003eChoi S-H, Oh Y-K, Koo D, editors. Visualization Techniques for Outlier Data. 2020 International Conference on Artificial Intelligence in Information and Communication (ICAIIC); 2020: IEEE.\u003c/li\u003e\n\u003cli\u003eStańczyk U, Jain LC. Feature selection for data and pattern recognition: An introduction: Springer; 2015.\u003c/li\u003e\n\u003cli\u003eZheng A, Casari A. Feature engineering for machine learning: principles and techniques for data scientists: \u0026quot; O\u0026apos;Reilly Media, Inc.\u0026quot;; 2018.\u003c/li\u003e\n\u003cli\u003eAl-Shehari T, Alsowail RA. An insider data leakage detection using one-hot encoding, synthetic minority oversampling and machine learning techniques. Entropy. 2021;23(10):1258.\u003c/li\u003e\n\u003cli\u003eKim M, Goh G. A sparse empirical Bayes approach to high‐dimensional Gaussian process‐based varying coefficient models. Stat. 2024;13(2):e678.\u003c/li\u003e\n\u003cli\u003eTalekar B, Agrawal S. A detailed review on decision tree and random forest. Biosci Biotechnol Res Commun. 2020;13(14):245-8.\u003c/li\u003e\n\u003cli\u003eZhang S. Challenges in KNN classification. IEEE Transactions on Knowledge and Data Engineering. 2021;34(10):4663-75.\u003c/li\u003e\n\u003cli\u003eSzeghalmy S, Fazekas A. A comparative study of the use of stratified cross-validation and distribution-balanced stratified cross-validation in imbalanced learning. Sensors. 2023;23(4):2333.\u003c/li\u003e\n\u003cli\u003eChen H, Wang N, Du X, Mei K, Zhou Y, Cai G. Classification prediction of breast cancer based on machine learning. Computational intelligence and neuroscience. 2023;2023(1):6530719.\u003c/li\u003e\n\u003cli\u003eCarrington AM, Manuel DG, Fieguth PW, Ramsay T, Osmani V, Wernly B, et al. Deep ROC analysis and AUC as balanced average accuracy, for improved classifier selection, audit and explanation. IEEE Transactions on Pattern Analysis and Machine Intelligence. 2022;45(1):329-41.\u003c/li\u003e\n\u003cli\u003eBrownlee J. Data preparation for machine learning: data cleaning, feature selection, and data transforms in Python: Machine Learning Mastery; 2020.\u003c/li\u003e\n\u003cli\u003eMerrick L, Taly A, editors. The explanation game: Explaining machine learning models using shapley values. Machine Learning and Knowledge Extraction: 4th IFIP TC 5, TC 12, WG 84, WG 89, WG 129 International Cross-Domain Conference, CD-MAKE 2020, Dublin, Ireland, August 25\u0026ndash;28, 2020, Proceedings 4; 2020: Springer.\u003c/li\u003e\n\u003cli\u003eLi X, Zhang H, Zhao S, Tang K. Predicting risky sexual behavior among college students through machine learning approaches: cross-sectional analysis of individual data from 1264 universities in 31 provinces in China. JMIR public health and surveillance. 2023;9(1):e41162.\u003c/li\u003e\n\u003cli\u003ePilehvari A, Chipoletti A, Krukowski R, Little M. Unveiling socioeconomic disparities in maternal smoking during pregnancy: a comprehensive analysis of rural and Appalachian areas in Virginia utilizing the multi-dimensional YOST index. BMC Pregnancy and Childbirth. 2024;24(1):828.\u003c/li\u003e\n\u003cli\u003eHtet AS, Bjertness MB, Sherpa LY, Kj\u0026oslash;llesdal MK, Oo WM, Meyer HE, et al. Urban-rural differences in the prevalence of non-communicable diseases risk factors among 25\u0026ndash;74 years old citizens in Yangon Region, Myanmar: a cross sectional study. BMC Public Health. 2016;16(1):1225.\u003c/li\u003e\n\u003cli\u003eNyirenda HT, Mulenga D, Silitongo M, Nyirenda HB, Nyirenda T. Spatial distribution and correlates of smoking in Zambia. BMJ open. 2019;9(8):e030044.\u003c/li\u003e\n\u003cli\u003eKoblinsky M, Anwar I, Mridha MK, Chowdhury ME, Botlero R. Reducing maternal mortality and improving maternal health: Bangladesh and MDG 5. J Health Popul Nutr. 2008;26(3):280-94.\u003c/li\u003e\n\u003cli\u003eDennis CL, Brennenstuhl S, Brown HK, Bell RC, Marini F, Birken CS. High-risk health behaviours of pregnancy-planning women and men: Is there a need for preconception care? Midwifery. 2022;106:103244.\u003c/li\u003e\n\u003cli\u003eRaru TB, Ayana GM, Zakaria HF, Merga BT. Association of higher educational attainment on antenatal care utilization among pregnant women in east africa using Demographic and Health Surveys (DHS) from 2010 to 2018: a multilevel analysis. International journal of women\u0026apos;s health. 2022:67-77.\u003c/li\u003e\n\u003cli\u003eLange AE, Mahlo-Nguyen J, Pierdant G, Allenberg H, Heckmann M, Ittermann T. Antenatal Care and Health Behavior of Pregnant Women-An Evaluation of the Survey of Neonates in Pomerania. Children (Basel). 2023;10(4).\u003c/li\u003e\n\u003cli\u003eCortes ML, Louzado JA, Oliveira MG, Bezerra VM, Mistro S, Medeiros DS, et al. Association between perceived stress and health-risk behaviours in workers. Psychol Health Med. 2022;27(4):746-60.\u003c/li\u003e\n\u003cli\u003eTheuring S, Rubagumya K, Schumann H, Harms G, Rubaihayo J, Wanyenze R. Sexual Risk Behavior in HIV-Uninfected Pregnant Women in Western Uganda. Arch Sex Behav. 2021;50(7):3277-86.\u003c/li\u003e\n\u003cli\u003eSarofim M, Huilan X, Wen LM, Baur LA. Associations of Sociodemographic Factors With Clusters of Obesity-Risk Behaviours Among Pregnant Women in Australia. Aust N Z J Obstet Gynaecol. 2025.\u003c/li\u003e\n\u003cli\u003eMurakami K, Hashimoto H. Associations of education and income with heavy drinking and problem drinking among men: evidence from a population-based study in Japan. BMC Public Health. 2019;19(1):420.\u003c/li\u003e\n\u003cli\u003eKebede SA, Weldesenbet AB, Tusa BS. Spatial Distribution and Factors Associated With Risky Health Behavior Among Adult Males Aged 15-59 Years in Ethiopia: Generalized Structural Equation Modeling. Front Psychiatry. 2022;13:688336.\u003c/li\u003e\n\u003cli\u003eDavis AM, Wambach KA, Nelson EL, Odar C, Lillis T, McKinley A, et al. Health behavior change in pregnant women: a two-phase study. Telemed J E Health. 2014;20(12):1165-9.\u003c/li\u003e\n\u003cli\u003eRamadani RC, Ibrahim K, Mirwanti R, Maulana S, Jabareen R. Social media use, knowledge, attitudes, and risky sexual behavior of HIV transmission: A survey among boarding school adolescent students in Indonesia. Belitung Nurs J. 2024;10(3):322-31.\u003c/li\u003e\n\u003cli\u003eTafti SP, Azarshab A, Mahmoudian RA, Khayami R, Abadi RNS, Ahmadi-Simab S, et al. Predictors of tobacco use among pregnant women: a large-scale, retrospective study. BMC Public Health. 2024;24(1):2544.\u003c/li\u003e\n\u003cli\u003eMirzakhani K, Shoorab NJ, Akbari A, Khadivzadeh T. High-risk pregnant women\u0026apos;s experiences of the receiving prenatal care in COVID-19 pandemic: a qualitative study. BMC Pregnancy Childbirth. 2022;22(1):363.\u003c/li\u003e\n\u003cli\u003eRockliffe L, Peters S, Heazell AEP, Smith DM. Factors influencing health behaviour change during pregnancy: a systematic review and meta-synthesis. Health Psychol Rev. 2021;15(4):613-32.\u003c/li\u003e\n\u003cli\u003eDowne S, Finlayson K, Tun\u0026ccedil;alp \u0026Ouml;, G\u0026uuml;lmezoglu AM. Provision and uptake of routine antenatal services: a qualitative evidence synthesis. Cochrane Database Syst Rev. 2019;6(6):Cd012392.\u003c/li\u003e\n\u003cli\u003eAkalu TY, Baraki AG, Wolde HF, Lakew AM, Gonete KA. Factors affecting current khat chewing among male adults 15\u0026ndash;59 years in Ethiopia, 2016: a multi-level analysis from Ethiopian Demographic Health Survey. BMC psychiatry. 2020;20:1-8.\u003c/li\u003e\n\u003cli\u003eRawlings G, Williams R, Clarke D, English C, Fitzsimons C, Holloway I, et al. Exploring adults\u0026rsquo; experiences of sedentary behaviour and participation in non-workplace interventions designed to reduce sedentary behaviour: a thematic synthesis of qualitative studies. BMC public health. 2019;19:1-16.\u003c/li\u003e\n\u003cli\u003eShahrabi Farahani F, Khosrowabadi R, Jaafari G. Risk-taking Behavior Under the Effect of Emotional Stimuli Among Children and Adults. Basic Clin Neurosci. 2022;13(4):585-93.\u003c/li\u003e\n\u003cli\u003eKonlan P, Ganle JK. Transactional sex and associated factors among young women in a tertiary institution in Northern Ghana: evidence from a cross-sectional survey. BMC Womens Health. 2025;25(1):298.\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"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":"bmc-pregnancy-and-childbirth","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"prch","sideBox":"Learn more about [BMC Pregnancy and Childbirth](http://bmcpregnancychildbirth.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/prch/default.aspx","title":"BMC Pregnancy and Childbirth","twitterHandle":"@BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Machine learning, Risky behavior, Light GBM, pregnant women, Shapley Additive Explanation (SHAP)","lastPublishedDoi":"10.21203/rs.3.rs-7364399/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7364399/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cstrong\u003eIntroduction\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eUnhealthy consumption patterns of substances, sexual activity, and physical inactivity are key contributors to morbidity and mortality for pregnant women. However, there is a limited study on those risk behaviors and their determinants among pregnant women in East Africa. Therefore, this study aimed to determine risky behaviors and their determinants among pregnant women in East Africa by using data from the DHS using machine learning algorithms.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMethods\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study utilized DHS data from 2012–2022 in 12 East African countries. Data was analyzed using Python version 3.7 and R version 4.3.3 for data preprocessing, modeling, and statistical analysis. Model performance was evaluated using accuracy and Area Under the Curve (AUC). Finally, the SHAP was applied in Python to further explore and interpret the predictors of risky behaviors among pregnant women aged 15–59 years old.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eResults\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eIn this study, the Light Gradient Boosting Machine model achieved an accuracy of 95.88% and an AUC score of 0.991. The SHapley Additive exPlanations analysis revealed that pregnant women who lived in rural areas, women with poor wealth income, women with middle wealth income, women whose husbands had primary education, and women not exposed to media increased risky behavior. Whereas women who were employed, women’s utilized ANC services, and women aged 25–36 lower likelihood of risky behaviors.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConclusion\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe Light GBM was the best-performing model for identifying determinants of risky behaviors among pregnant women in Eastern African countries. Interventions should focus on promoting and strengthening women’s ANC accessibility, improving husbands’ education, expanding media use, and economic empowerment for women to reduce the burden of risky behaviors.\u003c/p\u003e","manuscriptTitle":"Exploring Machine learning Algorithms to Identify Determinants of Risky Behavior among Pregnant Women 15–59 years in Eastern African countries using the Demographic and Health Survey data","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-09-15 06:43:45","doi":"10.21203/rs.3.rs-7364399/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-09-29T03:41:58+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-09-28T03:36:25+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-09-19T07:12:32+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"10745928215633132489901084356313734234","date":"2025-09-14T05:51:38+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-09-13T17:00:47+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"49337012473957830103661861692917116809","date":"2025-09-09T06:58:11+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"11744877825062605440572749518093984808","date":"2025-09-07T16:03:06+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"188960580837344177382398984397998466649","date":"2025-09-07T16:02:22+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"262925093858736540034144556924033578296","date":"2025-09-05T17:31:09+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"219940831610300725208713664782483973167","date":"2025-09-05T16:07:05+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"65718130511268063210199304544302160236","date":"2025-09-05T16:07:04+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-09-05T16:00:26+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-08-14T00:45:47+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-08-14T00:45:34+00:00","index":"","fulltext":""},{"type":"submitted","content":"BMC Pregnancy and Childbirth","date":"2025-08-13T11:08:24+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"bmc-pregnancy-and-childbirth","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"prch","sideBox":"Learn more about [BMC Pregnancy and Childbirth](http://bmcpregnancychildbirth.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/prch/default.aspx","title":"BMC Pregnancy and Childbirth","twitterHandle":"@BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"c11cc017-28ea-49f5-a50f-2a81335573c4","owner":[],"postedDate":"September 15th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2025-11-26T08:53:24+00:00","versionOfRecord":[],"versionCreatedAt":"2025-09-15 06:43:45","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7364399","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7364399","identity":"rs-7364399","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}