Application of the Decision Tree (C4.5) Algorithm for Classifying Elementary School Students' Interests

Abstract

Identifying elementary school students' interests is important so that teachers and schools can provide learning and extracurricular activities aligned with students' potential. This study applies the Decision Tree C4.5 algorithm to classify student interests into seven classes: Academic, Art, Sports, Technology, Language, Religious, and Leadership Interest. The study uses a 100-student synthetic dataset constructed from class-based profiles, with academic scores, classroom activity, extracurricular activities, and hobbies as input attributes. C4.5 was selected because the study emphasizes interpretable decision rules rather than claiming predictive superiority over ensemble methods. The dataset was divided into 80 training and 20 testing records using stratified sampling. The experiment produced 19 correct predictions, yielding 95% accuracy, with macro precision of 95.24%, macro recall of 95.24%, and macro F1-score of 94.29%. However, because the testing sample is small and the synthetic data contain deliberately structured relationships between inputs and target classes, the result cannot be generalized as C4.5 performance in real schools. The contribution of this study is a transparent methodological baseline and decision rules that can be traced by teachers. Validation using anonymized real student data, cross-validation, and empirical comparison with Random Forest and XGBoost are required in future research.