Comparative Analysis of Heterogeneous Ensemble Learning using Feature Selection Techniques for Predicting Academic Performance of Students

Abstract

Data stored in digital form is increasing daily, and so its complexity. Processing a massive volume of data needs efficient technology. Data mining and Machine Learning researchers are focused on finding a suitable algorithm that can find important information after processing that data. In educational data mining, most of the students' records are also stored in digital form. So, the researchers are also trying to find some informative knowledge that can be helpful for the students, teachers, and management to improve their working towards the success of the students and institution also. In predictive modelling, the main challenge is finding the most effective predictive techniques that help achieve an acceptable accuracy level. This article, therefore, proposes a hybrid or heterogeneous approach of Correlation Attribute Evaluation, Ensemble Learning like Stacking, Voting and MultiScheme, in conjunction with seven different Machine Learning algorithms to improve the prediction accuracy up to an acceptable level. Here, k-fold cross-validation was used as a test method to evaluate the predictive performance of the classification algorithms.