Klasifikasi Debitur Kartu Kredit Menggunakan Algoritme K-Nearest Neighbor untuk Kasus Imbalanced Data

Abstract

Data is said to suffer the class imbalanced problem when the class distribution are highly imbalance. In this case, minority class is more difficult to predict then the majority class. Though the minority class sometime has important information. In this paper, classification analysis of credit card debtors is conducted by using k-nearest neighbor that can classify debtors into two categories, good or bad. Analysis of a prospective debtor is essential to minimize credit risk. One approach taken to overcome imbalanced data problems is to modify instance distribution using oversampling and undersampling method. The evaluation is conducted by comparing the value of parameter k, accuracy, precision, recall, and F-measure. The evaluation results show that oversampling technique gives the best result of 96.24% with k = 3, 99.23% recall with k = 2, 95.21% precision with k = 1, and 96.30% F-measure with k = 3.