Student’s Academic Performance Prediction Using Ensemble Methods Through Educational Data Mining

Abstract

Student academic performance (SAP) is one of the most important factors that influence the attainment of an educational institution. Educational data mining (EDM) follows the principle of the data mining process that can be applied to educational data to analyze the student’s academic performance. In this paper, a student’s performance prediction model is proposed that includes new features such as the grade of the students and backlog information which is identified as the most important features based on the data mining method. In this study, we train and build models on a small size dataset and provide the feasibility of creating a prediction model which gives a satisfying accuracy rate. We also find the key attributes such as subject backlog, subject grade pay average which are essential for building models and visualization of data. The key attributes are fed into different algorithms to find the best model. The efficiency of the proposed model is calculated using different classifiers such as decision tree, k-nearest neighbor (KNN), and support vector machine (SVM). Furthermore, to enhance the performance of the classifier techniques, the ensemble methods like bagging, boosting are applied. The outcome of this study proves that there exists a direct correlation between the grades and their academic performance. The outcome of the study proves the efficiency of ensembled methods by providing an accuracy of 88.3%.