Modeling Student Behavior and Predicting Academic Performance Using LMS Data
6th International Interdisciplinary Symposium on Chaos and Complex Systems, SCCS 2025, İstanbul, Türkiye, 8 - 10 Mayıs 2025, ss.337-346, (Tam Metin Bildiri)
- Yayın Türü: Bildiri / Tam Metin Bildiri
- Doi Numarası: 10.1007/978-3-032-09101-7_29
- Basıldığı Şehir: İstanbul
- Basıldığı Ülke: Türkiye
- Sayfa Sayıları: ss.337-346
- Anahtar Kelimeler: Academic performance prediction, Educational Process Mining, Learning Management Systems, Machine learning, Student behavior analysis
- İstanbul Kültür Üniversitesi Adresli: Evet
Özet
The COVID-19 pandemic has significantly accelerated the adoption of digital education platforms, particularly Learning Management Systems (LMS), thereby amplifying the importance of behavior-driven analytics in education. While traditional Educational Data Mining (EDM) approaches primarily rely on frequency-based analysis of student interactions, this study integrates Process Mining (PM) techniques to model learning pathways and applies machine learning algorithms to predict academic performance. In this research, process models were derived from LMS log data collected from a university. Process-based indicators—most notably the fitness score—were extracted and combined with conventional EDM features to construct enhanced datasets. These enriched datasets were used to train predictive models using Random Forest, XGBoost, and Support Vector Machine, and their performance was benchmarked against models built solely on EDM features. The results indicate that incorporating process-oriented variables improved key classification metrics, including F1-score and ROC-AUC, by 2–4%. Additionally, SHAP analysis revealed that the fitness score was the most influential feature in predicting student success. By integrating procedural learning behavior into predictive analytics, this study advances the current EDM paradigm. The findings contribute to the development of explainable artificial intelligence in education, enabling more effective learning analytics, personalized interventions, and data-driven decision support systems in online learning environments.