Educational Data Mining Techniques for Behavioral Pattern Analysis

Abstract

Educational data mining (EDM) extracts actionable insights from learner interaction logs, assessment records, and platform engagement data to understand and predict student behavioural patterns -- how students study, when they disengage, which learning strategies correlate with success, and how peer interactions shape academic trajectories. Behavioural pattern analysis moves beyond outcome prediction to reveal the underlying learning processes that drive performance differences, enabling interventions targeted at behaviour change rather than outcome remediation. This paper proposes the Educational Behavioural Pattern Mining Framework (EBPMF), a systematic evaluation of six EDM techniques -- sequence pattern mining, clustering, association rule learning, process mining, temporal pattern analysis, and social network analysis -- across 12 behavioural pattern benchmarks in online and blended learning environments. EBPMF analyses 96,400 learner interaction logs across four platforms (Moodle, Coursera, edX, and a custom blended learning LMS) over 18 months. Key results: process mining achieves the highest Behavioural Insight Score (BIS = 0.912), revealing 8 distinct learning strategy archetypes; temporal pattern analysis identifies procrastination signatures predicting dropout with AUC = 0.906; social network analysis identifies peer influence clusters explaining 28.4% of grade variance. The framework provides EDM technique selection guidance and an open behavioural analytics benchmark suite.