Identifying Student Profiles within Online Judge systems using Explainable ArtificialIntelligence

Abstract

The rapid growth of online programming platforms such as Codeforces, LeetCode, and HackerRank has generated large volumes of student interaction data, creating new opportunities for intelligent learning analytics. This paper presents an approach for identifying student profiles within online judge systems using Explainable Artificial Intelligence (XAI), aiming not only to classify learners based on their coding behavior but also to provide transparent and interpretable insights into their performance. The proposed system collects and processes data such as problem-solving accuracy, submission attempts, execution time, difficulty levels, and temporal activity patterns to construct meaningful feature representations of students. Machine learning models, including classification and clustering techniques, are employed to categorize students into distinct profiles such as beginners, intermediate learners, advanced coders, and struggling participants. To address the limitations of traditional black-box models, the system integrates explainability techniques such as feature importance analysis, SHAP (Shapley Additive Explanations), and LIME (Local Interpretable Model-Agnostic Explanations), enabling educators and learners to understand the key factors influencing each prediction. The resulting framework not only enhances trust and interpretability in AI-driven educational systems but also supports personalized learning, targeted feedback, and performance improvement strategies. Experimental results demonstrate that the proposed approach effectively identifies diverse learning patterns while maintaining high accuracy and interpretability, making it a valuable tool for modern e-learning environments and intelligent tutoring systems.