ACADEMIC PERFORMANCE PREDICTION MODEL FOR STUDENTS BASED ON EDUCATIONAL DATA MINING WITH INTEGRATION OF LEARNING STYLE, SOCIAL, ECONOMIC, AND DEMOGRAPHIC FACTORS

Abstract

The challenge of homogeneous teaching in higher education often ignores student diversity, which hinders the realization of inclusive education. This study aims to develop an academic performance prediction model based on Educational Data Mining (EDM) by integrating 27 comprehensive features, including learning styles (VARK), social, economic, and demographic factors. Data from 266 students in the Department of Electrical Engineering and Informatics, Faculty of Engineering, State University of Malang, were analyzed using three algorithms, namely Decision Tree, Random Forest, and CatBoost, with the SMOTE technique applied to handle class imbalance. The results of the study indicate CatBoost as the superior model with an accuracy of 87.04%. Determinant factor analysis revealed that behavioral factors such as duration of hanging out with friends, duration of cell phone use, and psychological learning styles (VARK) were highly significant predictors. These findings confirm the importance of non-academic factors and learning styles, and imply the urgency of developing early detection systems and more personalized and adaptive teaching strategies.