Enhancing Predictive Accuracy in Education: A Detailed Analysis of Student Performance Using Machine Learning Models

Abstract

Predicting student performance is crucial for enhancing educational outcomes and providing targeted interventions. This study employs various machine learning models to comprehensively analyze student performance and enhance predictive accuracy in educational settings. It addresses the challenge of predicting student outcomes by considering diverse predictors such as demographic details, academic history, and behavioral factors. The aim of this research is to develop robust models capable of generalizing across different educational environments and providing actionable insights into factors influencing academic achievement.  The research utilized various approaches, including Random Forest, Linear Regression, Decision Tree, Support Vector Machine, Logistic Regression, XGboost, K-Nearest Neighbor, and Multilayer Perceptron Classifier. Results revealed significant outperformance of these traditional statistical methods. Attendance, previous academic performance, and engagement level emerge as critical predictors of success. Using the Student Attendance Dataset from Kaggle, Logistic Regression achieved an accuracy of 98%, Random Forest 100%, Multilayer Perceptron 100%, and XGBoost 96%. Findings underscore the potential of machine learning models as proactive tools for educators to effectively address student needs. The research highlights the transformative impact of machine learning in educational assessment and intervention strategies, advocating for a more data-driven approach in education. Future research may refine models by incorporating additional features, ultimately offering educators effective tools for supporting students and enhancing academic outcomes.