Data-Driven Mental Health Assessment of College Students Using ES-ANN and LOF Algorithms During Public Health Events

Abstract

Psychological stress in college students has attracted much attention due to its effects on psychological conditions during public health events. Traditional data mining can only get limited data through questionnaires, which is far from enough to present the whole situation for non-participants of the stress dynamic. This paper presents an advanced computational methodology for assessing psychological stress using data mining techniques. Herein, an ES-ANN model is developed to address the problem of imbalanced sample data. Besides, the LOF algorithm was realized, which allowed the comparison of the proposed approaches, including supervised and unsupervised learning methods applied in anomaly detection. Then, the results of extensive performance evaluation for the proposed ES-ANN model are performed by applying well-known G-mean and F1 score performance metrics. The results indicated that the ES-ANN model outperformed state-of-the-art benchmark methods, namely Random Forest and Decision Tree, with an increment of 8% in G-mean and an increment of 4% in F1 score. It proves the dependability and accuracy of the proposed ES-ANN model for the identification of students with high levels of psychological stress. Implementation of an integrated supervised-unsupervised learning system (ES-ANN and LOF) opens a new avenue to the early detection of psychological stress among college students