Preventing Student’s Mental Health Problems with the Help of Data Mining

Abstract

The increasing incidence of mental health issues among university students has become a significant concern, often referred to as a "mental health crisis" in academic settings. This study addresses the challenge of predicting mental health issues in university students using data mining techniques. The research involved the creation of a new dataset via a survey method focused on university students, covering various factors like behavioral traits, health conditions, and lifestyle choices. Data mining algorithms such as Naive Bayes, Random Forest, SVM, KNN, and Decision Tree were employed to predict mental health status. The study included dataset collection, cleaning, integration, transformation, reduction, discretization, and the use of Weka and Orange for data analysis. Therefore, exploratory analysis revealed that 53.4% of students reported depression, with a higher incidence among male students and those less involved in extracurricular activities. Predictive analysis showed Naive Bayes as the most accurate algorithm (65.91%) for this prediction task, followed by Random Forest, SVM, KNN, and Decision Tree. The performance was evaluated using accuracy, F1-Score, precision, recall, AUC, and CA. The study highlights the correlation between various aspects of university students' lives and depression. Active participation in extracurricular activities was found to lower depression risks. The effectiveness of data mining in understanding and predicting mental health issues in university students was established, with Naive Bayes being the most effective algorithm for this purpose.