Klasifikasi Debitur Kartu Kredit dengan Pemilihan Fitur Menggunakan Voting Feature Intervals 5

Abstract

Provision of credit cards for customers is one of the ways to obtain profit in banking activities which cause risks of losses if the customer frequently delinquent the payments. Therefore, it is important to know the banking profile of the customer who will apply for a credit card. The banking profile data is used as input for Voting Feature Intervals 5 (VFI5) algorithm in the development of classification models that aim to classify potential debtor based on the payment status of the debtor. The debtor data used in this research is categorized as imbalanced data, hence it is necessary to have other performance measures beside accuracy; in this research we also used recall and precision. The input data consist of 14 features, however each features has different significance in classifying debtor. Therefore a feature selection process is conducted before the development of the model. The feature selection is conducted using two approaches: feature selection based on the accuracy of each feature and stepwise feature selection. The former method provides the better accuracy of 67.74%, and the values of recall and precision for the class of bad debtor are 46.88% and 24.69%, respectively.