Clustering-Based EMT Model for Predicting Student Performance

Predicting students’ performance has emerged as an attractive task among researchers. They use supervised and unsupervised educational data mining (EDM) techniques to build an understandable and effective model. This helps decision makers enhance the performance of the students. The challenge of finding an optimal model leads to appearance of many techniques from both EDM techniques. Hence, we propose a unified framework to build a novel supervised cluster-based (CB) classifier model. The unified framework uses clustering technique to group historical records of students into a set of homogeneous clusters. Then, classifier model for each cluster is built and the final unified classifiers along with the centroids at each cluster are used as CB classifier model. The experimental results show that the CB model gains a high accuracy performance reached 96.25%. In addition, we use feature selection techniques for selecting the relevant features from a space of features. The model obtains a high accuracy performance using relevant features reached to 96.96% where the percentage of relevant features on average is 57.4% of overall features.