Predictive Modeling of Undergraduate Admissions in Bangladesh: A Comparative Analysis of Public and Private Universities Using Machine Learning and Explainable AI Techniques

The higher education landscape in Bangladesh has witnessed unbelievable amount of growth throughout the years yet remains a challenging domain for students due to intense competition and socio-structural disparities. This paper explores a dataset collected from 15 public and private universities across Bangladesh and investigates the key components which influences undergraduate admissions in public and private universities. Moreover, statistical analysis is elaborately used to uncover patterns in academic inclinations, behavioural trends and develop predictive models. Afterwards a correlation heatmap was created to explore the interdependent relationships between the features. The machine learning models were trained with the correlated features achieving estimated prediction for XGBoost, Random Forest and Support Vector Machine (SVM). Using these models, we obtained an accuracy of 85.83% for Random Forest, whereas we obtained 85% for both XGBoost and Support Vector Machine (SVM). Furthermore, to enhance interpretability, explainable AI techniques such as SHAP (Shapley Additive Explanation) and LIME (Local Interpretable Model-agnostic Explanations) were applied to obtain a better visualization of the predicted models. To conclude, this study presents a comprehensive analysis highlighting the role of academic performance, socioeconomic background and behavioural aspects of a student's undergraduate admission journey through predictive modelling and explainable AI techniques. Further research would require a more diverse and better dataset with more features of behavioural trends of undergraduate students from Bangladesh.