Measuring Student Digital Behavior and Academic Performance Using Decision Support System

Abstract

This study proposes a decision support system for measuring student digital behavior and predicting academic performance using machine learning algorithms. The system analyzes digital activity features, including login frequency, access duration, assignment submission patterns, and learning interaction, to classify students into high-performance, moderate, and at-risk categories. This study implements and compares Random Forest, Support Vector Machine, and Logistic Regression models to identify the most effective predictive approach. We obtain that the Random Forest model achieves the best performance with an accuracy of 0.89, demonstrating superior capability in handling complex and non-linear behavioral data. The selected model is integrated into the proposed decision support system and applied to 200 student records, producing classification results of 31% high-performance students, 49% moderate, and 20% at-risk. Furthermore, the system generates academic recommendations based on the prediction outcomes to support monitoring and early intervention strategies. Feature importance analysis reveals that assignment submission is the most influential factor in predicting academic performance, followed by login frequency and access duration. This study demonstrates that integrating artificial intelligence with decision support systems can produce reliable predictive insights and improve the effectiveness of academic monitoring and educational decision-making.