Impact of Residency Status on Academic Performance of Bangladeshi University Students Using Machine Learning and Explainable AI

This research investigates the effect of residency status (residential or commuting) on the academic achievement, as measured by CGPA, of students in Bangladesh's universities. Past studies have largely been unable to control for the difference of the residency status from other variables that affect it, including participation, attendance, and study habits. With the application of PSM and machine learning methods, this research tried to fill this gap by examining the exclusive effect of residency status controlling for such confounding factors. The total sample of 387 students from 10 universities was examined, and some machine learning models, i.e., Random Forest and Gradient Boosting, were used to predict CGPA. The average accuracy of the best performing model was 83%. The results show that the residency status did not have any quantifiable impact on CGPA when study habits were held constant, whereas LIME analysis reaffirmed that study habits and attendance were the all-encompassing determinants, as opposed to type of housing. The result suggests that students can buffer the impact of living conditions by maintaining good academic habits.