Predicting Engineering Students’ Grade on Introductory Physics Using Machine Learning

Abstract

Introductory physics is a compulsory course for the first-year engineering college for providing students the underlying concepts of the future course throughout their study. A sudden shift of distance learning during the disruption of COVID-19 in the middle of 2021 has generated an extensive collection of educational data that can potentially be mined for educational purposes. Educational data mining (EDM), a branch of machine learning research, has offered some tools to perform this task. In this study, a logistic regression classifier was employed to early identify students' performance in introductory physics courses for engineering majors. Data were collected at a public university (N=180) through a learning management system engaged throughout a semester. This study successfully trained the model with an 80% identification rate to predict the low-performing student in the course. The finding is necessary for the educator to do the review and give feedback to their class for providing some help, particularly for the low-performing student. It is suggested for the further development of the model to make prediction more accurate with another model ensemble that has been proven decisive in the recent study of machine learning.