Abstract
Background Depression is a common mental disorder, and prior research has primarily focused on changes in depression among college students based on single psychological factors (such as personality traits or social support); there is currently a lack of research on predicting changes in depression based on multiple factors. We observed changes in symptoms of depression among college freshmen after enrollment and applied machine learning (ML) methods to longitudinally and comprehensively investigate personality traits, family factors, and social factors as determinants of changes in depression among college students. Methods We studied 5,534 college freshmen and measured personality traits, family factors, social factors, and other variables of college students twice. We used four ML algorithms—random forest (RF), support vector machines (SVM), logistic regression (LR), and k-nearest neighbors (KNN)—to predict changes in depression among college students. Results The ML algorithms demonstrated reasonable performance in the prediction task, with the non-linear kernel SVM performing the best (averaging 95% accuracy). Additionally, we found that neuroticism, positive coping, psychoticism, extraversion, and maternal autonomy were the most influential features in our study for classifier prediction. Conclusions Our results suggest that applying ML methods to study changes in depression among college students may be feasible, as personality traits appear to predict changes in depression among college students and may be suitable for screening for prevention interventions.