Evaluating the Effectiveness of Machine Learning Models in Predicting Student Academic Achievement

Abstract

Student performance is influenced by various factors, including demographic, psychological, social, and institutional determinants. This study comprehensively analyzes these factors to understand their relative impacts on academic achievement. Data from 1000 students, including variables such as gender, ethnicity, parental education level, lunch type, test preparation course participation, and scores in mathematics, reading, and writing, were analyzed. Categorical variables were converted to numerical values and normalized, followed by exploratory data analysis and correlation analysis. Machine learning models, including Linear Regression, Decision Tree Regressor, Random Forest Regressor, K-Nearest Neighbors Regressor, Support Vector Regressor (SVR), Gradient Boosting Regressor (GBR), XGBoost Regressor, and Neural Network (MLP), were used to predict student performance. The study found that test preparation courses significantly enhance student performance, with strong positive correlations between mathematics, reading, and writing scores. Demographic factors such as gender, ethnicity, parental education level, and lunch type were not direct determinants of student performance. The SVR and Linear Regression models exhibited the best predictive performance. These findings highlight the need for optimized educational strategies focusing on effective test preparation and the interconnectedness of subject performance to enhance academic achievement.