Predicting engineering students' academic performance using ensemble classifiers- a preliminary finding / A’zraa Afhzan Ab Rahim and Norlida Buniyamin

• Main Authors: Ab Rahim, A’zraa Afhzan, Buniyamin, Norlida
• Format: Article
• Language: English
• Published: UiTM Press 2022
• Subjects: Performance. Competence. Academic achievement; Prediction analysis
• Online Access: https://ir.uitm.edu.my/id/eprint/63176/1/63176.pdf; https://doi.org/10.24191/jeesr.v20i1.013; https://ir.uitm.edu.my/id/eprint/63176/; https://jeesr.uitm.edu.my/v1/; https://doi.org/10.24191/jeesr.v20i1.013

Current literature review indicates a void of an accurate predictive tool to assist educators and administrators in analyzing and monitoring student performance in Malaysia. Wellknown data mining classifiers such as Decision Tree (DT), Support Vector Machine (SVM), Logistic Regression (LR), Naïve Bayes (NB), and K-nearest neighbor (KNN) have been traditionally used but often suffer from the high variance and overfitting issues indicated by good performance on training data but relatively poor on unseen data. To address these problems, alternative ensemble classifiers such as Extreme Gradient Boosting (XGB), Random Forest (RF), and Heterogeneous Ensemble Method (HEM) are evaluated/proposed. This paper aims to compare the performance of single versus ensemble classifiers. In addition, another overarching research objective is to predict students' CGPA during their final semester grades by augmenting the more widely used cognitive with non-cognitive features to obtain a holistic solution. Not only will the accuracy among classifiers be compared, but another priority measure is their recall value to ensure each sample is classified correctly. It is found that ensemble classifiers outperform their single classifiers in terms of both accuracy and recall. Preliminary results indicate that augmenting cognitive features with non-cognitive features results in better accuracy in classifiers and can classify samples according to their respective classes with less variability.