A Novel Approach for University Admission Prediction using Feature selection and Data Upgrading Strategies

Abstract

The uncertainty surrounding university admissions has become a significant concern for students aspiring to secure positions in reputable academic institutions. With increasing competition and complex admission criteria, applicants often struggle to evaluate their chances accurately. This paper presents a robust machine learning-based approach to predict the probability of university admissions by leveraging student academic profiles and entrance exam scores. The core objective is to provide a data-driven framework that can assist both students in making informed decisions and institutions in streamlining the selection process. The study begins by collecting and preparing a diverse dataset that includes students' GRE, TOEFL, EAMCET, NEET, and PGCET exam scores, undergraduate GPA, university ratings, and categorical demographic attributes such as gender and caste—crucial in regional contexts where reservation policies are in place. Multiple supervised machine learning models were developed and compared, including Linear Regression, Decision Tree, Random Forest, K-Nearest Neighbors (KNN), Support Vector Machines (SVM), and LASSO regression. Feature selection techniques were employed using methods such as best subset selection and decision tree feature importance, while hyperparameter tuning was conducted using Grid Search to improve model accuracy. Evaluation metrics such as Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE) were used to measure and compare the performance of the models. Among all the approaches, the Random Forest algorithm demonstrated superior accuracy and robustness in predicting admission chances.