Predicting Emotional Dysregulation Using Machine Learning: The Role of Digital Rumination, Sleep Variability, Cognitive Load Reactivity, and Social Micro-Withdrawal

Objective: The present study aimed to develop and validate a machine learning model to predict emotional dysregulation based on digital rumination, sleep variability, cognitive load reactivity, and social micro-withdrawal among young adults.Methods and Materials: This quantitative predictive study was conducted on a sample of ۴۲۸ young adults in Japan using a multimodal data collection approach integrating self-report measures, ecological momentary assessment, and passive digital tracking over a six-week period. Emotional dysregulation was assessed using a validated scale, while predictor variables included digital rumination, sleep variability derived from wearable actigraphy, cognitive load reactivity measured through task-based and self-report indices, and social micro-withdrawal operationalized via behavioral smartphone data. Data preprocessing involved normalization, missing data imputation, and feature engineering. Multiple machine learning algorithms, including random forest, support vector machine, gradient boosting, and XGBoost, were trained and evaluated using a ۷۰-۱۵-۱۵ data split. Model performance was assessed using RMSE, MAE, and R², and interpretability was enhanced through SHAP analysis to determine feature importance and interaction effects.Findings: The results indicated that the XGBoost model achieved the highest predictive accuracy (R² = ۰.۷۶), outperforming other algorithms. Digital rumination emerged as the strongest predictor, followed by social micro-withdrawal, cognitive load reactivity, and sleep variability. Significant interaction effects were observed, particularly between digital rumination and social micro-withdrawal, as well as between digital rumination and sleep variability, indicating synergistic influences on emotional dysregulation. SHAP analysis revealed nonlinear relationships and threshold effects, demonstrating that higher levels of digital rumination and behavioral disengagement substantially increased the likelihood of elevated emotional dysregulation.Conclusion: The findings highlight the effectiveness of machine learning approaches in capturing the complex and interactive determinants of emotional dysregulation, emphasizing the central role of digitally mediated cognitive and behavioral processes alongside physiological variability. Integrating digital rumination, sleep patterns, cognitive load sensitivity, and micro-level social behaviors provides a comprehensive framework for understanding emotional dysregulation in contemporary contexts and offers valuable implications for targeted, data-driven mental health interventions. Objective: The present study aimed to develop and validate a machine learning model to predict emotional dysregulation based on digital rumination, sleep variability, cognitive load reactivity, and social micro-withdrawal among young adults. Methods and Materials: This quantitative predictive study was conducted on a sample of ۴۲۸ young adults in Japan using a multimodal data collection approach integrating self-report measures, ecological momentary assessment, and passive digital tracking over a six-week period. Emotional dysregulation was assessed using a validated scale, while predictor variables included digital rumination, sleep variability derived from wearable actigraphy, cognitive load reactivity measured through task-based and self-report indices, and social micro-withdrawal operationalized via behavioral smartphone data. Data preprocessing involved normalization, missing data imputation, and feature engineering. Multiple machine learning algorithms, including random forest, support vector machine, gradient boosting, and XGBoost, were trained and evaluated using a ۷۰-۱۵-۱۵ data split. Model performance was assessed using RMSE, MAE, and R², and interpretability was enhanced through SHAP analysis to determine feature importance and interaction effects. Findings: The results indicated that the XGBoost model achieved the highest predictive accuracy (R² = ۰.۷۶), outperforming other algorithms. Digital rumination emerged as the strongest predictor, followed by social micro-withdrawal, cognitive load reactivity, and sleep variability. Significant interaction effects were observed, particularly between digital rumination and social micro-withdrawal, as well as between digital rumination and sleep variability, indicating synergistic influences on emotional dysregulation. SHAP analysis revealed nonlinear relationships and threshold effects, demonstrating that higher levels of digital rumination and behavioral disengagement substantially increased the likelihood of elevated emotional dysregulation. Conclusion: The findings highlight the effectiveness of machine learning approaches in capturing the complex and interactive determinants of emotional dysregulation, emphasizing the central role of digitally mediated cognitive and behavioral processes alongside physiological variability. Integrating digital rumination, sleep patterns, cognitive load sensitivity, and micro-level social behaviors provides a comprehensive framework for understanding emotional dysregulation in contemporary contexts and offers valuable implications for targeted, data-driven mental health interventions.

کلیدواژه ها:

Emotional Dysregulation ، Machine Learning ، Digital Rumination ، Sleep Variability ، Cognitive Load Reactivity ، Social Micro-Withdrawal