Pinpointing Key Success Factors in Bangladesh’s Public University Entrance Exams: A Feature-Optimized SVM Architecture with XAI
Anwar Hossain Efat
Abstract
Each year, over a million candidates vie for limited spots in entrance exams across 53 public universities in Bangladesh, with roughly 36 candidates per seat. Although there are 112 private universities, public institutions remain the preferred choice for many. This intense competition often leads to unfulfilled aspirations and psychological distress. Factors behind unsuccessful admission attempts include the limited number of seats, varied educational and familial backgrounds, distractions during preparation, unsuitable activities during preparation, and socio-economic challenges. This study aims to identify these key factors and assess their impact using data from 600 students at 15 public and private universities. Employing Machine Learning techniques, the research focuses on predicting admission out-comes. The proposed ‘Feature-optimized Explainable Support Vector Machine (FoX-SVM)’ model achieves a notable 95.88% accuracy in predicting public university admissions, aided by hyperparameter tuning. The study also utilizes Explainable Artificial Intelligence (XAI) to enhance model interpretability, identifying crucial factors influencing admission success. Refining the model by removing the least important features demonstrates how focusing on key factors improves performance and guides future candidates more effectively.