Uji Akurasi Algoritme K-Nearest Neighbor Dan Naïve Bayes Dalam Klasifikasi Kelayakan Pemberian Kredit Perbankan

Abstract

Banking credit is a process of giving money or debt following an agreement between the borrower and the bank, as well as determining the classification of creditworthiness on housing loans (KPR). This affects the customer's waiting time for the results of a bank's decision, the success of a bank's credit management will greatly affect the fate of many customer funds if the analysis is inaccurate, so technology is needed to find hidden information on prospective borrowers' data to predict a customer's loan repayment ability. This study uses an algorithm K-Nearest Neighbor and Naïve Bayes to determine the eligibility classification for bank lending and determine the accuracy of bank credit eligibility for mortgages, to determine the accuracy of the algorithm through three stages of testing, namely several preprocessing stages starting from checking duplicates, deal with missing value, deal with outliers, do label encoding, deal with data imbalance use method SMOTE, and standardize using scaler standard. The results of the Naïve Bayes and KNN algorithms as well as the model stages are evaluated to examine each stage in the data for the model's ability to predict, the evaluation matrix used is in the form of a results confusion matrix. There is the best result, namely the KNN algorithm in the third test with a value of K = 10 with a performance of 80.92% training data accuracy and 78.86% testing data and getting a score confusion matrix TP 76 and TN 21.