A Multi-Method Statistical and Machine Learning Analysis of Living Arrangements and Student Mental Well-being

Abstract

Mental health among students is an emerging issue in the tertiary education sector, as it relies on academic, social and environmental factors. This study examines the potential effects of living arrangements on student mental health using a multi-method analytical framework. The recommended approach includes the use of descriptive statistics, inferential test, Structural Equation Modeling (SEM), Bayesian inference, time series analysis, and machine learning techniques. Results have shown statistically significant differences in living conditions (p < 0.05) such that living independently is associated with higher anxiety and social isolation. SEM shows social isolation (= -0.52) to be the most predictive of mental well-being and the second strongest predictor, then financial stress (= -0.36). Accuracy of machine learning models up to 88 bets on the credibility of forecasts. The framework offers a data-driven and scalable method of comprehending, anticipating, and enhancing student mental health The Random Forest model achieved a predictive accuracy of 88.3%, demonstrating strong predictive performance for student mental well-being classification.