Analysis of Online Learning Behavior in Java Programming Design Courses using a Hybrid Machine Learning Approach

With the increasing integration of digital platforms in computer science education, understanding student behavior in online Java programming design courses has become essential for improving engagement and learning outcomes. Specifically, analyzing behavioral patterns in Java programming courses can reveal critical insights into learning styles, performance gaps, and content interaction. However, traditional monitoring techniques often lack the scalability and analytical depth needed to uncover hidden behavioral structures. To address these challenges, this study presents a hybrid Machine Learning (ML) approach that integrates K-Means clustering with a Boosted Neural Network (BNN), implemented via AdaBoost-enhanced Extreme Learning Machine (ELM), to detect and classify learning behavior patterns among students enrolled in a Java programming course. The dataset used in this analysis was collected from a university-level online learning management system and includes detailed logs of student activity, such as code submissions, quiz attempts, video views, and forum participation. Initially, K-Means clustering is employed to group learners into meaningful behavioral clusters. These clusters are then used to train the BNN, which accurately classifies student behavior into categories such as consistent performers, late submitters, and passive participants. The classification model achieves an accuracy of 98.9%, precision of 98.7%, recall of 99.1%, specificity of 98.6%, and F1-score of 98.9%, confirming the effectiveness of the approach. These findings support the development of adaptive learning interventions and personalized content delivery in online programming education