Higher Education Hybrid Machine Learning Model for Assessing the Impact of Stress on Student Performance in Higher Education

Abstract

This study proposes a hybrid machine learning model to evaluate the impact of stress on student performance in higher education, integrating psychological, academic, and behavioral indicators into a unified predictive framework. Recognizing that stress is a multidimensional construct influenced by academic workload, social pressures, financial concerns, and personal wellbeing, the research combines traditional statistical analysis with advanced machine learning techniques to improve predictive accuracy and interpretability. The hybrid model leverages feature extraction through principal component analysis (PCA), stress-level classification using ensemble methods, and performance prediction through a stacked regression architecture. Data were collected from undergraduate and postgraduate students using standardized stress-assessment instruments, academic records, and digital learning activity logs. Experimental results demonstrate that the hybrid approach outperforms single-model baselines, achieving higher precision in identifying high-risk students and stronger correlations between predicted and actual academic outcomes. Furthermore, explainability techniques such as SHAP values reveal that time-management issues, sleep quality, assessment load, and emotional resilience are the most influential predictors. The findings highlight the potential of hybrid machine learning systems to support early-warning mechanisms and personalized interventions within higher education. This research contributes to the development of data-driven strategies that enhance student wellbeing and academic success by proactively addressing stress-related challenges.