Mining Student data by Ensemble Classification and Clustering for Profiling and Prediction of Student Academic Performance

Applying Data Mining (DM) in education is an emerging interdisciplinary research field also known as Educational Data Mining (EDM). Ensemble techniques have been successfully applied in the context of supervised learning to increase the accuracy and stability of prediction. In this paper, we present a hybrid procedure based on ensemble classification and clustering that enables academicians to firstly predict students’ academic performance and then place each student in a well-defined cluster for further advising. Additionally, it endows instructors an anticipated estimation of their students’ capabilities during team forming and in-class participation. For ensemble classification, we use multiple classifiers (Decision Trees-J48, Naïve Bayes and Random Forest) to improve the quality of student data by eliminating noisy instances, and hence improving predictive accuracy. We then use the approach of bootstrap (sampling with replacement) averaging, which consists of running k-means clustering algorithm to convergence of the training data and averaging similar cluster centroids to obtain a single model. We empirically compare our technique with other ensemble techniques on real world education datasets.