Towards the integration of multiple classifier pertaining to the Student's performance prediction

Abstract

Summary Accurate predictions of students’ academic performance at early stages of the degree programme helps in identification of the weak students and enable management to take the corrective actions to prevent them from failure. Existing single classifier based predictive modelling is not easily scalable from one context to another context, Moreover, a predictive model developed for a particular course at a particular institution may not be valid for a different course at the same institution or any other institution. With this necessity, the notion of the integrated multiple classifiers for the predictions of students’ academic performance is proposed in this article. The integrated classifier consists of three complementary algorithms, namely Decision Tree, K-Nearest Neighbour, and Aggregating One-Dependence Estimators (AODE). A product of probability combining rule is employed to integrate the multiple classifiers for the prediction of academic performance of the engineering students. This approach provides a generalized solution for student performance prediction. The proposed method has been applied and compared on three student performance datasets using t-test. The proposed method is also compared with KSTAR, OneR, ZeroR, Naive Bayes, and NB tree classifiers as well as with the individual classifiers.