Predicting students’ academic achievement in the digital age of education requires an understanding of how they interact with online learning environments. To leverage engagement measures using a graph analysis model, this study suggests a novel way to model and analyze students’ course performance. With the use of an online learning management system, we can create a graph-based depiction of student interactions and activities that helps us identify patterns of engagement that traditional learning accomplishment indicators might miss. To determine the level of learning achieved in each course, the graph analysis model in this study evaluated several features, including quiz scores, individual and group project completion rates, student participation in discussion boards, and attendance. Utilizing data from the Binus Online Learning Management System, our experimental findings show that the Orthogonal Matching Pursuit model is the most optimum regression model to predict the FIN feature using the given input dataset. The predicted FIN variable is then added into the input dataset which were used further as data variable to predict the PASS/FAIL variable using classifier models. Next, our experimental findings show that the Graph Similarity approach and GED distance function was the most optimum classifier model in predicting students’ PASS/FAIL in the mentioned course. This research highlights the potential of graph-based analysis to identify at-risk students and tailor personalized interventions to enhance learning experiences.