An Efficient 2-Stages Classification Model for Students Performance Prediction

Abstract

Students’ performance prediction is one of the essential educational data mining research fields. Predicting students’ performance aims at improving the learning process inside educational institutions. This is achieved by early prediction of at-risk students who are vulnerable to drop out to help them and improve their performance sooner. Therefore, several machine learning techniques were proposed to predict students’ performance. The aim of this research is to improve the prediction by proposing two 2-Stages classification models. Several machine learning techniques: Multilayer Perceptron (MLP), Decision Tree (J48), Random Forest (RF), Gradient Boosting Method (GBM) and Logistic Regression (LR), have been experimented as part of voting method. The classification models were trained and tested on a dataset generated from an e-learning system used in previous studies. The performance of the proposed models was evaluated not only using the accuracy measure, but also using F1-Score and Cohen’s kappa score measures. The best obtained result was with the proposed 2-Stages 3-Classifiers classification model implementing voting method. The classifiers of the first stage consist of MLP, J48, and RF. The classifiers of the second stage consist of J48, RF, and LR. The proposed 2-Stages 3-Classifiers classification model achieved an accuracy of 77.26%, F1-Score of 77.8%, and Cohen’s kappa score of 72.0%. The proposed 2-Stages 3-Classifiers classification models outperformed other classification models in predicting students’ performance.