Using Machine Learning to Construct a Simplified Scale for Assessing Mental Status Among Undergraduates

Due to pressures related to academic or emotional factors, undergraduates may face certain psychological risks. Early psychological risk screening is helpful in identifying mental health issues among undergraduates, enabling timely intervention or treatment. This study collected assessment data from 32,868 undergraduates using four scales (SBQ-R, PHQ-9, GAD-7, ISI) from three different universities in China. Among them, 2,632 representative undergraduates were selected for scale utility analysis to explore simplified screening approaches. In this paper, machine learning methods were employed, treating each item as a feature. The forward feature selection was conducted using the maximum relevance-minimum redundancy (mRMR) approach to identify key items capable of distinguishing different levels of psychological risk among undergraduates. The study evaluated the performance of four machine learning algorithms under varying numbers of items. Through this approach, the scale was simplified while maintaining assessment reliability, effectively reducing the time required for psychological evaluation.