PREDICTION OF COOPERATIVE LOAN FEASIBILITY USING THE K-NEAREST NEIGHBOR ALGORITHM

Abstract

Approval of credit lending to cooperative members without proper feasibility analysis can cause credit problems, cooperatives such as late payment of installments, and an increase in bad credit which can threaten the survival of the cooperative as a provider of lending services. As a solution to minimize the creditworthiness assessment errors for loan funds, research is carried out to analyze the feasibility of loan funds from the data of cooperative members using the data mining method approach and the algorithm used using the K-Nearest Neighbor. The purpose of this research is to predict the feasibility of granting credit with the right decision and to find out the level of evaluation, accuracy, and validation of the effectiveness of the k-NN algorithm on processing creditworthiness application data classifications. After the prediction research was carried out, the data on the eligibility of credit lending applications were conducted at the Bakti Berkah Sukaraja Savings and Loan Cooperative, The data obtained from the accuracy value of the k-nearest neighbor algorithm before being validated has an accuracy of 87.78% with AUC 0.95, after validation with split validation the accuracy decreased slightly by 2% to be 85.71%, while the AUC value in the ROC Curve was 0.836%. Even though there was a decline, it can still be categorized as a good classification. The impact of this research is that besides the accuracy of the k-NN algorithm being validated, the Bakti Berkah Sukaraja Savings and Loan cooperative can predict the feasibility of applying for credit funds, as an effort to reduce the threat of bad credit risk

Keywords

• accuracy
• creditworthiness
• data mining classification
• k-nearest neighbor
• roc curve split validation