A Machine Learning Approach for Predicting Academic Performance Using Classification Models

Abstract

This study predicts academic performance using data from 1,100 students across private universities in Bangladesh, comparing seven machine-learning models: K-Nearest Neighbor, Decision Tree, Gaussian Naive Bayes, Support Vector Machine, Logistic Regression, Random Forest, and Linear Regression, across five performance categories (excellent, satisfactory, moderate, unsatisfactory, and probation). The Decision Tree was the most accurate model at 97.10%. Social media usage and probation status showed significant negative correlations, while SGPA, CGPA, attendance, and study hours were positive predictors. The findings offer actionable insights for early intervention and personalized academic support, with deep-learning-based evaluation planned as future work.