Perbandingan Algoritma C4.5 dan CART dalam Memprediksi Kategori Indeks Prestasi Mahasiswa

Abstract

This research compared the accuracy of prediction of Grade Point Average (GPA) of the first semester students using C4.5 and CART algorithms in Faculty of Information Technology (FTI), Universitas Kristen Duta Wacana (UKDW). This research also explored various parameters such as numeric attribute categorization, data balance, GPA categories number, and different attributes availability due to the difference of data availability between Achievement Admission (AA) and Regular Admission (RA). The training data used to create decision tree were FTI students, 2008-2015 batch, while the testing data were FTI students, 2016 batch. The accuracy of prediction was measured by using crosstab table. In AA, the accuracy of both algorithms can be achieved about 86.86%. Meanwhile, in RA the accuracy of C4.5 is about 61.54% and CART is about 63.16%. From these accuracy result, both algorithms are better to predict AA rather than RA.