Article

Uncovering Lifestyle and Mental Well-being Predictors of Academic Performance Change in Online Learning: A Comparative Analysis of Interpretable Machine Learning Models

---

Abstract

The transition to online learning has reshaped academic engagement and well-being among students, yet the factors driving changes in performance remain poorly understood. This study investigates how lifestyle and mental health indicators predict self-reported changes in academic performance during online learning using interpretable machine learning models. A dataset of 1,000 students was analyzed through a comparative framework employing Logistic Regression and Random Forest classifiers, complemented by SHAP-based explanations. Descriptive analysis revealed balanced demographic distributions, with most students reporting moderate stress levels and similar proportions across performance categories. Model results showed comparable accuracies of approximately 0.33, reflecting the complexity of predicting academic outcomes. However, both models consistently identified screen time, sleep duration, and physical activity as the most influential predictors, while stress level and exam anxiety exhibited smaller yet coherent effects. Logistic Regression highlighted categorical distinctions such as education level and anxiety, whereas Random Forest captured nonlinear interactions among lifestyle variables. SHAP analyses provided global and local interpretability, confirming that higher screen exposure reduced the likelihood of improvement, while adequate sleep and regular physical activity were positively associated with better outcomes. These findings emphasize the central role of lifestyle balance in sustaining academic performance and mental well-being during remote education. Despite modest predictive power, the interpretable modeling approach offers actionable insights for educators, policymakers, and students to foster healthier and more effective online learning environments.

Keywords: Online Learning, Lifestyle Factors, Student Mental Health, Interpretable Machine Learning, Academic Performance

How to Cite:

Fortuna, D. & Hutagalung, C., (2025) “Uncovering Lifestyle and Mental Well-being Predictors of Academic Performance Change in Online Learning: A Comparative Analysis of Interpretable Machine Learning Models”, Artificial Intelligence in Learning 1(4), 271-286. doi: https://doi.org/10.63913/ail.v1i4.41