Comparison analysis of data mining methodology and student performance improvement influence factors in small data set

Abstract

Based on Programme for International Student Assessment survey, Indonesia student performance was on the lower position compared to other participated countries. Nevertheless, the actual reason of someone's performance in studying is hard to predict. Therefore, we need to limit the research's scope for finding more specific influencing factors. In this study, a group of students who have the same learning method, teacher, course, and also facility in a learning environment is observed to find significant influencing factors. We develop a questionnaire with various factors that are related to students' characteristic. It is administered to students of Junior High School Muhammadiyah 2 Depok Sleman in the same year. Consequently, data gathered is in small size. In order to yield maximum accuracy in small dataset, SMOTE is used to generate new data synthetically hence instance number is increasing. Besides several approaches were analyzed by using combination of preprocessing and variation feature selections. The result of this research shows that attribute subset selection by Classifier Subset Evaluator (CSE) yields the best result based on Naive Bayes accuracy and variance. Various significant factors influencing studying performance of tested students were also found including blood type, who student live with, father's education, mother's education, kind of activity done in spare time and favorite course.