Multi-timepoint clustering analysis of self-regulated learning profiles based on user interaction logs in a virtual learning environment

With the growing use of Virtual Learning Environments in education, understanding how students regulate their learning has become increasingly important. This study analyzes data from a course delivered through a platform designed with a pedagogical architecture to support self-regulated learning. The data were extracted from Moodle logs using Structured Query Language (SQL) queries on the system’s database, in the context of a seven-week “Introduction to Python Language” course offered at a Brazilian public educational institution. For the analysis, we applied both k-means and Agglomerative clustering algorithms with a fixed and parsimonious two-cluster solution ( \(k = 2\) ) to identify binary self-regulated learning profiles, characterize learning patterns, and examine their relationship with students’ academic performance based on cluster analyses conducted at multiple points of the course. The Agglomerative clustering algorithm proved effective for the multi-point analysis of clusters, enabling a clear examination of variations in learning profiles across course weeks. Our findings indicate that students with higher levels of self-regulated learning tend to engage more consistently with learning resources and achieve better academic performance. Statistically significant differences between the identified profiles were observed across multiple course points, supported by nonparametric hypothesis testing and multiple comparison procedures. Overall, this study demonstrates the utility of Educational Data Mining techniques in profiling student learning behaviors. The results provide valuable insights into how learners manage their educational processes, supporting the design of pedagogical strategies that foster self-regulation, autonomy, and more effective learning experiences in virtual environments.