A Comparative Analysis of Explainable Artificial Intelligence (XAI)-Based Machine Learning Models for Student Performance Prediction

Aktaş S.

ARCENG V. International Aegean Scientific Studies Congress, Muğla, Türkiye, 11 - 12 Temmuz 2026, ss.166-177, (Tam Metin Bildiri)

  • Yayın Türü: Bildiri / Tam Metin Bildiri
  • Basıldığı Şehir: Muğla
  • Basıldığı Ülke: Türkiye
  • Sayfa Sayıları: ss.166-177
  • Marmara Üniversitesi Adresli: Evet

Özet

Accurately and early predicting students’ academic performance is critical for the timely identification of at-risk students and the planning of targeted academic interventions. Although machine learning–based models achieve high predictive accuracy, the lack of transparency in their decision-making processes, particularly in human-centered domains such as education, constitutes a major limitation for trust and practical applicability. In this study, Logistic Regression, Decision Tree, Random Forest, Radial Basis Function kernel–based Support Vector Machines, and k-Nearest Neighbor algorithms are jointly evaluated for the student performance prediction task, and their decision mechanisms are examined using SHAP and LIME-based Explainable Artificial Intelligence methods. The experiments are conducted on the widely used Student Performance dataset, which comprises demographic, familial, and academic variables for 649 students and 33 attributes. The models are compared using accuracy, precision, recall, and F1-score metrics, and the results show that Logistic Regression and Random Forest achieve more balanced and superior classification performance than the other approaches. SHAP analyses indicate that prior academic achievement indicators and specific socio-academic factors are decisive in model predictions, while LIME-based local explanations enable reliable interpretation of individual student-level predictions. By jointly evaluating both classification performance and SHAP- and LIME-based interpretability within the same experimental setting, this study presents an integrated framework for developing transparent and trustworthy artificial intelligence-based decision support systems in education.