OPTIMASI KLASIFIKASI DATA KINERJA AKADEMIK MAHASISWA MENGGUNAKAN SVM BERBASIS ALGORITMA GENETIKA

Abstract

For a college, especially a private university, students are the main component that supports the survival of the college. An educational database containing information about students is useful for predicting student academic performance. Several studies on the classification of academic performance have been conducted, it is clear that classification problems generally exist in the number of attributes, too many unnecessary attributes will increase computational time and reduce accuracy. The combination of PSO + SVM has proven to be more effective than SVM in various types of datasets. Therefore, this study will try to compare SVM-GA for the classification of student academic performance so that students who have good and bad academic performance can be seen. The data used is the academic performance data of the midwifery students of Ngudi Waluyo University, 2012-2014. The highest accuracy of SVM-GA is the accuracy of 93.55% and AUC 0.977. The previous SVM method had an accuracy of 90.51% and AUC 0.963. Based on the AUC value, the performance of the proposed SVM-GA method is in the "Perfect" group.