Predicting Student's Board Examination Performance using Classification Algorithms

Abstract

Educational Data Mining can help stakeholders give appropriate decisions to improve educational experiences. New knowledge or models are realized when data mining techniques are applied on educational data. This study focuses on building classification model by utilizing data mining techniques for predicting the likelihood of a student to pass the Licensure Examination for Teachers (LET). Several well-known data mining algorithms such as Neural Network, Support Vector Machine, C4.5 Decision Tree, Naive Bayes, and Logistic Regression are used to build the models. The performance of these models in terms of accuracy to predict the student's performance in the Licensure Examination for Teachers, F1 Measure and Area under the Curve (AUC) value were compared to determine which among these classification algorithms performs best. Results show that C4.5 turns to be the most suitable algorithm for the model. It has an accuracy of 73.10%, F1 measure of 62.53% and Area under the Curve value of 0.730. The identified model could be able to identify students who will likely fail the Licensure Examination for Teachers. These students should be given higher priority during their mock board review and be able to pass the board examination. Aside from helping these students, it also helps the institution get higher percentage of passers in the Licensure Examination for Teachers and can be of help during accreditation.