EARLY DETECTION OF DEPRESSION AMONG STUDENTS USING AI- DRIVEN PREDICTIVE MODELS

Abstract

Depression among university students is a growing concern, significantly affecting academic performance, emotional well-being, and social interactions. Traditional methods for detecting depression, such as self-report questionnaires and clinical interviews, face challenges related to stigma and accessibility. In this study, we leverage machine learning techniques to develop an efficient predictive model for detecting depression in students based on behavioral, psychophysiological, and demographic factors. We applied multiple classification algorithms using a publicly available dataset, including Logistic Regression, Support Vector Machine (SVM), Random Forest, Decision Tree, K-nearest neighbors (K-NN), and Naïve Bayes. Performance evaluation was conducted using accuracy, precision, recall, and F1-score. Results indicate that Logistic Regression outperforms other models, achieving an accuracy of 92%, making it a reliable predictor of depression among students. Our findings highlight the potential of machine learning in mental health applications and suggest the need for AI-driven tools to support early intervention strategies in academic institutions. Future work should explore the integration of real-time data sources and deep learning techniques to further enhance prediction accuracy.