Optimized Machine Learning Framework for Predicting Student Mental Health Crises

Mental health is crucial for overall well-being, particularly among students, as delayed detection can lead to depression, anxiety, and even suicide. Early intervention is vital to prevent worsening conditions and provide timely support. This study predicts mental health outcomes among Bangladeshi university students using machine learning techniques. Data were collected from 314 students across 26 universities through a 66-question survey covering demographics, academics, family, and mental well-being. Multiple feature selection methods, combined with PCA and RFE-RF, identified the 14 most important features for evaluating six machine learning models. The Decision Tree classifier achieved the highest accuracy of 98.41%, surpassing prior approaches. This research underscores machine learning’s potential for early mental health assessment, paving the way for scalable, data-driven solutions to support students facing mental health challenges.