COMPARISON C4.5 AND NAÏVE BAYES METHODS BASED ON PARTICLE SWARM OPTIMIZATION IN LEVELS OF DROP OUT STUDENTS

Abstract

The high percentage of drop-out students causes a campus management problem, this is because the percentage of students graduating on time is one of the elements of accreditation assessment set by the national accreditation board of higher education. One reason why the drop out rate is still high is because the Management System has not run well, such as lecturer professionalism, campus facilities, academics and administration, student affairs, outside influence and student personality. This study aims to analyze several indicators that can cause student drop outs by comparing the C4.5 method based on particle swarm optimization and Naïve Bayes based on PSO. This study contributes to campus management in anticipating the occurrence of drop outs through indicators that occur and can predict student drop out rates through the classification process. The highest level of accuracy produced from C4.5 + PSO is around 99.32% with AUC from Naïve Bayes is 0.974 categorized as excellent classification.