Course performance prediction for basic courses of universities based on support vector machine

Abstract

As the popular area of educational data mining, course performance prediction is the prediction of the final performance of students in a course, assisting educators’ personalized teaching and reducing the pressure of students. With proper features and algorithms specified, prediction models with high accuracy can be built. Course performance prediction for basic courses of universities requires feature selection from both subjective and objective aspects. GPA, grades of prerequisite courses, assignment scores and the inquiry count are selected as the objective features and the individual interest is specified as the subjective feature. The output of the prediction is the students’ final performance divided into 5 levels. The Gaussian support vector machine, the polynomial support vector machine, BP neural network, random forest and logistic regression were employed as the classifier, with the accuracy and AP of the five algorithms compared. It is found that the Gaussian support vector machine combined with selected features can reach the optimal accuracy and AP, reaching 99%. With the Gaussian support vector machine applied, a course performance prediction model for basic courses of universities is proposed, which provides a novel method for the study on course performance prediction for basic courses of universities.