Perbandingan Metode k-Nearest Neighbor, Regresi Logistik Biner, dan Pohon Klasifikasi pada Analisis Kelayakan Pemberian Kredit

Abstract

Kredit Tanpa Anggunan (KTA) are bank loans given to debtors without asking for guarantees. Some debtors who have made KTA but still need additional loan funds can top up. However, offering this facility to the public cannot be separated from the risk that the debtor and/or other parties fail to fulfill their obligations to the bank. In an effort to assess the feasibility of prospective debtors, banks need decision-making tools so that they can easily and quickly estimate which debtors are able to pay off credit on time (good credit). The tool that is often used is classification. In this study, we will compare 3 classification methods, namely k-nearest neighbor, binary logistic regression, and classification tree, to obtain the best method for analyzing the feasibility of giving KTA top-up. Based on the accuracy value in each method, the classification tree produces the highest accuracy value compared to the other two methods. Thus, for this study, the classification tree is the best method, with an accuracy value of 87.68%. The variables used in the classification tree are DBR, length of service of a debtor, credit limit, type of debtor's occupation, the total income of the debtor, the area where the debtor lives, and the credit period of the debtor is 1 month.