A Comprehensive Evaluation of Machine Learning Techniques for Forecasting Student Academic Success

Abstract

Improving academic outcomes relies on accurately anticipating student outcomes within a course or program. This predictive capability empowers instructional leaders to optimize the allocation of resources and tailor instruction to meet individual student needs more effectively. In this study, we endeavor to delineate the attributes of machine learning algorithms that excel in forecasting student grades. Leveraging a comprehensive dataset encompassing both personal student information and corresponding grades, we embark on a rigorous evaluation of various regression algorithms. Our analysis encompasses a range of widely used technniques, Incorporating various machine learning algorithms like XGBoost, Linear Regression, K-Nearest Neighbor, Decision Tree, Random Forest, and Deep Neural Network. By conducting thorough comparisons using metrics such as Root Mean Squared Error, determination coefficient, Mean Average Error and Mean Squared Error. Our aim is to pinpoint the algorithm that exhibits superior predictive ability. Notably, our experimental findings unveil the deep neural network as the standout performer among the evaluated algorithms. Having an outstanding coefficient of determination of 99.95% and Minimal error margins, the DNN emerges as a potent tool for accurately forecasting student grades. This discovery not only underscores the efficacy of advanced machine learning methodologies but also underscores the transformative potential they hold in shaping educational practices and optimizing student outcomes.