Work in progress: Early prediction of students' academic performance in an introductory engineering course through different mathematical modeling techniques

Abstract

This paper presents our ongoing efforts in developing mathematical models to make early predictions, even before the semester starts, of what score a student will earn in the final comprehensive exam of the engineering dynamics course. A total of 1,938 data records were collected from 323 undergraduates in four semesters. Employed were four different mathematical modeling techniques: multivariate linear regression, multilayer perceptron neural networks, radial basis function neural networks, and support vector machines. The results show that within the five predictor variables investigated in this study, there is no significant difference in the prediction accuracy of these four mathematical models.