Early Detection of Student Mental Health and Academic Burnout Using Multimodal AI-Based Behavioral, Physiological, and Emotional Analysis

Abstract

Mental health issues such as stress, anxiety, depression, and academic burnout are increasingly common among university students and have a significant impact on academic performance and long-term well-being. Existing assessment approaches rely mainly on self-reported questionnaires and periodic evaluations, which are reactive, subjective, and ineffective for early intervention. This paper presents a multimodal artificial intelligence-based system for early identification of student mental health conditions by analyzing behavioral, physiological, emotional, and academic data. The proposed framework integrates facial emotion recognition, wearable sensor data analysis, natural language processing of reflective text to continuously monitor student well-being in a privacy-aware manner. Machine learning and deep learning models are employed to detect stress, anxiety, and burnout indicators and to predict future mental health risks. Experimental results obtained from real and synthetic datasets demonstrate that multimodal analysis provides more reliable and accurate predictions than single-source methods. The proposed system enables early risk identification and supports timely intervention in academic environments.