Klasifikasi Debitur Kartu Kredit Menggunakan Algoritme Support Vector Machine Linear Kernel untuk Kasus Imbalanced Data

Abstract

In this research, classification analysis of credit card debtors is conducted by using support vector machine linear kernel that can classify debtors into two categories good or bad. The data used in this research is imbalanced because most data are from one class. Classification algorithms generally result in poor performance on imbalanced data because the minority class is more difficult to predict than the majority class. One way that can be used to solve this problem is by using a sampling method with oversampling and undersampling technique. This research compares the value of accuracy, recall, precision, and F-measure. The evaluation result shows a fairly high accuracy values in the original data is 83.59% but, the value of recall, precision, and F-measure are 0%. Random oversampling technique gives the best performance with 54.14% accuracy, 53.47% recall, 61.30% precision, and 54.51% F-measure.