MOOCs generate high volumes of interaction data of the learner at a massive level, yet the challenges of how to sustain learner engagement and reduce the incidence of dropout rates remain unresolved by the educators due to the varying learning behaviours and evolving interests of the learners. To counter this limitation, the study suggests a hybrid rule-based and machine-learning framework, which will allow to establish a dynamic user profile in MOOC platforms by integrating both clickstream analysis and clustering learner trajectories. The framework is informed by applying explicit pedagogical rules, which classify learners into three levels of engagement, performance and risk, and simultaneously, using HDBSCAN-based clustering of trajectories to demonstrate the implicit temporal behavioral modes in the sequential interaction data. The research conducts experiment on the Open University Learning Analytics Dataset (OULAD) that comprises interaction records of 32,593 students enrolled in 22 courses. The results demonstrated that the hybrid method could succeed in categorizing the learners into meaningful behavioral categories such as steady, irregular, late starter and dropout-prone without using previous clusters. Clustering encourages a Silhouette Score of 0.306, Davies-Bouldin Index of 0.559 and Calinski-Harabasz Index of 1912.55, which suggest high cluster separation and stability. Furthermore, the visualization and correlation analyses confirm that the engagement features are the most prominent in differentiating the learners. The offered system can offer an interpretable, scalable, and flexible answer to learner profiling that can be used to support personalized recommendations, early warning, and interventions in the environment of large-scale MOOCs.
Xin Jin (2026) studied this question.