PREDICTING STUDENT ACADEMIC PERFORMANCE: A MACHINE LEARNING ANALYSIS OF STUDY HABITS AND LIFESTYLE FACTORS

Student academic performance prediction has become increasingly important for educational institutions seeking to implement targeted interventions and support systems.This study analyzes a comprehensive dataset of 1,000 university students to identify key factors influencing exam performance using advanced machine learning techniques. We examine the relationships between study habits, mental health, screen time, sleep patterns, and academic outcomes. Our analysis reveals that study hours per day exhibits the strongest correlation with exam scores (r = 0.825), followed by mental health rating (r = 0.322). Ridge regression achieved the highest predictive accuracy with an R score of 0.9015 and RMSE of 5.03 points. Feature engineering with polynomial terms significantly improved model performance, with study hour derivatives accounting for over 60% of predictive importance. Clustering analysis identified three distinct student performance groups with average scores of 61.95, 81.97, and 96.14 points respectively. Threshold analysis demonstrates that students studying 4+ hours daily score 35+ points higher than those studying less than 2 hours. Mental health ratings above 7 correlate with approximately 15-point score improvements. These findings provide actionable insights for educational interventions and student support systems.