Prediction of Risk for Health-Related Maladaptive Behaviors Using Hybrid Machine Learning Algorithms and Cognitive–Emotional Markers

Abstract

The present study aimed to predict the risk of health-related maladaptive behaviors based on cognitive–emotional markers using hybrid machine learning algorithms among university students. This applied study employed a descriptive–correlational predictive design. The statistical population consisted of university students in Karaj during the 2025–2026 academic year, from whom 420 participants were selected using multistage cluster sampling. Data were collected through standardized measures assessing maladaptive health behaviors, emotion regulation, cognitive impulsivity, stress sensitivity, cognitive flexibility, and lifestyle indicators. Data analysis was conducted using Support Vector Machine, Random Forest, Gradient Boosting, Artificial Neural Networks, and an Ensemble hybrid model. Model performance was evaluated using accuracy, sensitivity, specificity, and Area Under the ROC Curve indices. Inferential analyses indicated that cognitive–emotional markers significantly predicted maladaptive health behaviors. The hybrid ensemble model achieved the highest predictive accuracy (0.91) with an AUC value of 0.96, demonstrating excellent classification performance. Emotional dysfunction, cognitive impulsivity, and stress sensitivity emerged as the strongest predictors, whereas emotion regulation and cognitive flexibility acted as protective factors against maladaptive behavioral risk. The findings suggest that integrating cognitive–emotional indicators with hybrid machine learning algorithms provides an effective framework for early identification of individuals at risk for maladaptive health behaviors and supports the development of personalized preventive interventions.