Analisis Credit Scoring untuk Kelayakan Pemberian Kredit Usaha Rakyat (KUR) dengan Backpropagation Neural Network dan TOPSIS

Abstract

The People's Business Credit (KUR) scheme is a financing program designed to support the growth of Micro, Small, and Medium Enterprises (UMKM). To implement this, banks require a credit scoring system to assess the eligibility of prospective borrowers objectively. This study aims to analyze credit scoring for KUR eligibility using the Backpropagation Neural Network method and TOPSIS at Bank BRI Unit Tanggul. The BPNN method is employed to classify KUR applicants as eligible or ineligible, while TOPSIS is used to rank eligible borrowers using Rank Order Centroid (ROC) weighting. The dataset comprises information on 150 KUR applicants, obtained through interviews with staff at Bank BRI Unit Tanggul. Six attributes are utilized—income, business duration, residence status, collateral, loan amount, and loan term—with credit eligibility status serving as the target variable. Preprocessing steps include data cleaning, transforming categorical data into numerical data, and Min-Max normalization. The results indicate that the BPNN method achieved an accuracy of 93.33%, precision of 91.67%, recall of 100%, and an F1-Score of 96.65%. The 116 customers classified as eligible were subsequently ranked using TOPSIS. The ranking results show that the top 25 customers (ranked 1–25) possessed preference values ranging from 0.828642 to 0.507761, and all were approved for the loan. The study concludes that the combination of BPNN and TOPSIS effectively supports the credit-scoring process by determining eligibility and prioritizing KUR disbursement.