An Effectual Model for Early Prediction of Academic Perfomance using Ensemble Classification

Abstract

In the past few years, researchers are focused towards educational data mining (EDM) to improve the quality of education. Student’s academic performance prediction is a vital issue for improving the value of education. Research study conducted in the literature review mainly captivated the academic performance prediction at higher education. Though the academic performance at secondary level is infrequent, the same could be a scale for a student's performance at subsequent levels of education. Poor grades at lower levels also impact student’s future performance. In this paper, an effectual model is built with the help of significant factors that affect a student's academic performance at secondary level using single and ensemble techniques of classificification For this, both single and ensemble classification techniques are used in this paper To do the same, three single classifiers (classification algorithm) i.e., MLP, Random Forest and PART along with three well recognized ensemble algorithms Bagging (BAG), LogitBoost (LB) and Voting (VT) are applied on the datasets. For better performance of aforementioned classifiers, blended versions (single + ensemble-based classifiers) of classification models are also built. Assessment metrics i.e., accuracy, precision, recall and F-measure used to evaluate the performance of our proposed model. Evaluation results shows that Logitboost with Random Forest outperformed with 99.8% accuracy. It is clearly visible from results that the proposed model is useful for academic performance prediction to improve learning outcomes in future.