Students Academic Performance Prediction with k-Nearest Neighbor and C4.5 on SMOTE-balanced data

Abstract

Success in predicting student academic performance from an early age will make it easier for teachers to provide assistance to students who have academic abilities below the class average or who have difficulty following the learning process in the classroom. This study uses a public dataset to predict student academic performance based on a number of attributes that students have, both static and dynamic. This study compares the performance of two classifiers, namely C4.5 and k-Nearest Neighbor (KNN) and applies the SMOTE preprocessing method in the classification of student academic performance. Experiments carried out using the Rapid Miner application resulted in the fact that the C4.5 Decision Tree method resulted in better prediction performance in terms of accuracy, recall, and precision values, respectively 71.09%, 71.63%, 71.54% compared to the K-Nearest Neighbor algorithm.