Leveraging Process Mining and Machine Learning to Enhance Performance Analysis

Abstract E-learning is a powerful means of promoting educational equity, blending modern pedagogical strategies with advanced information technologies. As its adoption continues to grow, optimising and enhancing e-learning processes becomes critical to ensuring effective, high-quality learning experiences. This paper presents a two-layered approach combining Process Mining and Machine Learning techniques to improve performance analysis in e-learning environments. The first layer involves applying process mining tools (Disco and ProM) to event logs extracted from a CS1 MOOC, in order to discover business process models and analyse process performance. The second layer consists of building a predictive model using five machine learning algorithms (Decision Tree, Linear Regression, Logistic Regression, Random Forest, and SVM) on the dataset generated in the first step. Our results reveal that SVM and Decision Tree models provide the most reliable performance predictions, effectively distinguishing between successful and unsuccessful students. Finally, we derive actionable recommendations based on behavioural indicators such as submission frequency, text insertion, and focus patterns, to help instructors and platform designers improve learning experiences.