Machine Learning: Driven Strategies for Enhancing Student Academic Performance

In order to improve learning outcomes and reduce dropouts, educational institutions must anticipate students' academic performance. This study focuses on identifying effective prediction models and employing a wrapper feature selection method to estimate students' performance in final examinations. Utilizing a well-established dataset from diverse engineering departments, comprising 3560 data samples, the research aims to select the optimal algorithm and feature selection technique for predicting the number of course arrears for each student. Regression predictive modeling, specifically for numerical value prediction, is employed along with relevant metrics to assess model performance. The study employs various machine learning approaches to forecast students' academic performance based on acquired data, including ongoing assessments and examinations. A comparative evaluation of ML techniques using different metrics is presented. This information empowers students to monitor their academic progress and adapt their study strategies for improved performance. Additionally, five feature selection strategies are applied to identify the most accurate predictor. The study incorporates regression algorithms and wrapper feature selection methods to enhance student performance outcomes. Experimental results highlight the effectiveness of the Variance Inflation Factor (VIF) with adaptive Decision Tree regression as the superior approach.