Students Performance in Board Examination Analysis using Naïve Bayes and C4.5 Algorithms

Abstract

© 2020, World Academy of Research in Science and Engineering. All rights reserved. In this generation, where the educational system is being added with new policies and requirements, the higher education institutes (HEIs) use data mining tools and methods for scholastic improvement of student performance. The authors gathered information from the Bachelor of Secondary Education (BSEd) graduates of Davao del Norte State College (DNSC), thru the Institute of Education (IEd) and other offices concerned. The information is comprised of six (6) attributes for analysis that are as follows: stanine, general weighted average (GWA), honors received, the type of scholarship that the specified respondents have, review center admission, and Licensure Examination for Teachers (LET) remarks. After the tools were performed on the datasets, it is found out that the general weighted average (GWA) is the most influential with the attribute's value of 0.171 and followed by review center admission with the attribute's value of 0.132. The performance of the models Naïve Bayes and C4.5 in predicting the students’ performance in the Licensure Examination for Teachers were compared in terms of accuracy to identify which among the two classification algorithms performs best. Results show that Naïve Bayes has 85.14% and 90% prediction accuracy using both 10-fold cross-validation and 70% training and 30% testing, while C4.5 accumulated 86.13% and 86.66% accuracy for 10-fold cross-validation and 70-30 percentage split, respectively. Using Naive Bayes algorithm, the AUC results to 0.8507 and 0.8333 when using the C4.5 algorithm. Therefore, it is construed that Naive Bayes results are more consistent, suitable and reliable when testing the data compared to C4.5. Recommendations cited in this study as to expanding the precision of the prediction of student performance in future.