Data Mining Approach to Predict Success of Secondary School Students: A Saudi Arabian Case Study

TLDR: In this article , three models were constructed using different algorithms: Naïve Bayes (NB), Random Forest (RF), and J48, which achieved a prediction accuracy of 99.34%.

Abstract: A problem that pervades throughout students’ careers is their poor performance in high school. Predicting students’ academic performance helps educational institutions in many ways. Knowing and identifying the factors that can affect the academic performance of students at the beginning of the thread can help educational institutions achieve their educational goals by providing support to students earlier. The aim of this study was to predict the achievement of early secondary students. Two sets of data were used for high school students who graduated from the Al-Baha region in the Kingdom of Saudi Arabia. In this study, three models were constructed using different algorithms: Naïve Bayes (NB), Random Forest (RF), and J48. Moreover, the Synthetic Minority Oversampling Technique (SMOTE) technique was applied to balance the data and extract features using the correlation coefficient. The performance of the prediction models has also been validated using 10-fold cross-validation and direct partition in addition to various performance evaluation metrics: accuracy curve, true positive (TP) rate, false positive (FP) rate, accuracy, recall, F-Measurement, and receiver operating characteristic (ROC) curve. The NB model achieved a prediction accuracy of 99.34%, followed by the RF model with 98.7%.