Detecting Depression in College Students Through Social Media Text Mining: A Predictive Modelling Approach

ASM Sc. J., 21(2), 2026

Published on September 15, 2026

https://doi.org/10.32802/asmscj.2026.0002

Author: Chee Keong Ch’ng, Pui Xian Chung

Abstract

Depression is a rising mental health concern among college students, often manifesting through linguistic and behavioural cues on social media platforms such as Facebook. This study aims to analyse social media comments using text mining techniques to detect potential signs of depression. The research applies a structured methodology beginning with text preprocessing, word cloud visualisation, and content analysis to uncover patterns indicative of emotional distress. Hierarchical clustering is then used to group related depressive themes. Several predictive models, including Logistic Regression, Support Vector Machine (SVM), Decision Tree, Random Forest, and Naive Bayes, are developed and evaluated using performance metrics such as Area Under the Curve (AUC), Classification Accuracy (CA), Precision (Prec), F1-score, Recall, and Matthews Correlation Coefficient (MCC). Based on the evaluation results, logistic regression outperformed the other methods and was selected for deployment to predict future instances of depression from social media text. The findings provide valuable insights into students’ digital expressions and support the development of early intervention tools for mental health monitoring.

Keywords: content analysis, depression detection, predictive model, social media, text mining