Code repository supporting the analysis of learning trajectories in software engineering education under AI policy changes

This repository provides a reproducible analysis pipeline for educational data mining research. It processes Moodle Learning Management System (LMS) event logs to identify student learning behavior patterns and evaluate their association with academic performance in software engineering courses. The pipeline consists of three Jupyter notebooks: Data pseudonymization and processing (A1): merges multi-year Moodle activity logs with student evaluation records while preserving privacy through salted SHA-256 cryptographic hashing. Implements fuzzy name matching to link records across datasets with inconsistent formatting. Event log analysis (A2): parses Moodle event descriptions to extract course module interactions, infers resource availability dates from teacher activity, and constructs student×resource interaction matrices. Includes an interactive Panel dashboard for exploring engagement patterns across milestones. Behavioral clustering and statistical validation (A3): applies K-Means clustering on PCA-reduced behavioral features to identify student engagement profiles. Validates clusters using MANOVA, univariate ANOVA with partial η² effect sizes, and Tukey HSD post-hoc tests. Includes cross-cohort t-tests comparing academic performance between year groups.