Predicting Stress among Students via Psychometric Assessments and Machine Learning

Sagnik Ghosh, Kirti Tripathi, Ankita Garg, Dinesh Singh, Amit Prasad, Arnav Bhavsar, Varun Dutt

Abstract: Yoga is recommended as a method for managing stress among students. Despite its widespread acceptance, however, there exists a gap in research concerning the prediction of changes in psychological stress among students practicing yoga. The main objective of this study was to assess the immediate consequences of practicing yoga among students on their psychological stress levels and to predict changes in stress levels via machine learning. A total of 166 participants were recruited in this study and were randomly divided into two groups: intervention (N = 110) and control N =56). The intervention group engaged in regular, structured sessions of the common yoga protocol by the Indian Government thrice a week in 45 minutes sessions over a period of 6 weeks. The self-reported questionnaire such as the Depression, Anxiety, Stress Scales (DASS 21), was employed to assess the psychological stress change before and after the intervention. Additionally, machine learning (ML) algorithms such as Linear Regression (LR), Random Forest (RF), Support Vector Regression (SVR), and Artificial Neural Networks (ANN) were developed to predict changes in stress levels as a result of the intervention from data collected before the intervention. These models were evaluated on metrics such as R-squared and root mean square percent error (RMSPE), with the RF algorithm showing the lowest RMSPE of 1.23 units and R-squared of 0.71 by relying on top 7 features. This research not only affirms the positive effects of yoga on psychological health but also highlights the utility of machine learning in predicting stress changes, offering new perspectives on yoga's role in stress management.