An SVM approach to predict student performance in manufacturing processes course

Abstract

Predicting student performance in core engineering courses is an important and challenging problem. The results of these predictions can be used to develop effective teaching strategies to improve learning and increase understanding. In this paper, the predictions of data mining models developed using Support Vector Machines (SVM) and Artificial Neural Networks (ANN) on student performance in the core engineering course Manufacturing Processes are compared with the predictions of the traditional multivariate statistical approach Multivariate Linear Regression (MLR). The predictor variables include a student's GPA and scores in six prerequisite courses. Comparisons are based on 504 data records collected from 63 students in five semesters. The results show that the predictions of SVM models outperform the predictions of an MLR model and an ANN model; and considering the grades of all the prerequisite courses is a need to predict performance of a student in a core course.