Artificial Intelligence-Based Multimodal Student Monitoring for Early Risk Detection

Abstract

Student monitoring in elementary education remains challenging when academic records, attendance, activeness, and emotional indicators are managed through fragmented administrative processes. This condition limits teachers’ ability to identify student risks early and provide timely interventions. This study aimed to develop an artificial intelligence-based multimodal student monitoring system for early risk detection by integrating academic grades, attendance records, facial expression data, and student activeness into a single web-based platform. The system was developed using the Waterfall model, comprising requirements analysis, system design, implementation, testing, deployment, and maintenance. Requirements were collected through observation and semi-structured interviews. The implemented system included student data management, manual and camera-based attendance tracking, academic score processing, transcript generation, academic progress analysis, risk classification, and intervention recommendations. Functional validation used black-box testing principles, and system performance was compared with the manual method. The system implemented monitoring functions. Black-box testing showed that login, score input, attendance input, wellness analysis, data editing, data export, and graph visualization produced expected outputs. Compared with the manual method, the system improved recording speed from 3.7 to 4.2, data neatness and accuracy from 3.7 to 4.4, report access from 4.0 to 4.6, and overall performance from 4.0 to 4.2. The developed system provides a decision-support tool that improves administrative efficiency and supports the detection of student risk. However, further evaluation is needed to measure recognition accuracy, scalability, usability, and predictive validity.