Authors: Vanaja Kumari Degala
Abstract: Early prediction of student academic performance has become an essential research problem in higher education due to increasing dropout rates and declining academic outcomes. The ability to identify at-risk students at an early stage enables institutions to implement timely interventions and personalized academic support. With the rapid growth of educational data, machine learning (ML) techniques have shown significant potential in extracting meaningful patterns from student records. This paper presents a comprehensive machine learning-based framework for early prediction of student academic performance using pre-admission data and first-year academic attributes. Several supervised learning algorithms, including Logistic Regression, Support Vector Machine, Random Forest, K-Nearest Neighbors, and Extreme Gradient Boosting (XGBoost), are evaluated. Dimensionality reduction using t-distributed Stochastic Neighbor Embedding (t-SNE) is employed to visualize high-dimensional student data. Experimental results demonstrate that combining admission scores with first-year course performance significantly improves prediction accuracy. The proposed approach can assist academic institutions in proactive decision-making to enhance student success and retention.
