Regression models of predicting student academic performance in an engineering dynamics course

Abstract

Prediction of student academic performance helps instructors develop a good understanding of how well or how poorly the students in their classes will perform, so instructors can take proactive measures to improve student learning. Based on a total of 2,151 data points collected from 239 undergraduate students in three semesters, a new set of multivariate linear regression models are developed in the present study to predict student academic performance in Engineering Dynamics - a high-enrollment, high-impact, and core engineering course that almost every mechanical or civil engineering student is required to take. The inputs (predictor/independent variables) of the models include a student's cumulative GPA; grades earned in four prerequisite courses: Engineering Statics, Calculus I, Calculus II, and Physics; as well as scores earned in three Dynamics mid-exams. The output (outcome/dependent variable) of the models is a student's final exam score in the Dynamics course. Multiple criteria are employed to evaluate and validate the predictive models, including R-square, shrinkage, the average prediction accuracy, and the percentage of good predictions. A good prediction is defined as the one with the prediction error of ±10%. The results show that the developed predictive models have the average prediction accuracy of 86.8%-90.7% and generate good predictions of 44.4%-65.6%. The implications of the research findings from the present study are also discussed. © American Society for Engineering Education, 2010.