Analysis of Data Mining Applications for Determining Credit Eligibility Using Classification Algorithms C4.5, Naïve Bayes, K-NN, and Random Forest

Abstract

This study aims to enhance the credit evaluation process within Credit Union (CU) Karya Bersama Lestari (KABARI). The study leveraged four distinct algorithms, namely Decision Tree C4.5, Naive Bayes, K-Nearest Neighbors (K-NN), and Random Forest, to predict the suitability of extending loans to potential borrowers. Rapid Miner was employed as a tool to maximize accuracy by analyzing the Confusion matrix. Testing was conducted on a dataset consisting of 459 member loan submissions. The results of the analysis revealed that the K-Nearest Neighbors (K-NN) algorithm achieved the highest accuracy among the evaluated algorithms. Specifically, the Decision Tree algorithm demonstrated an accuracy rate of 95.65%, along with a precision and recall of 94.12%. The Naive Bayes algorithm achieved an accuracy rate of 95.65%, supported by precision and recall values of 100% and 88.24%, respectively. The K-Nearest Neighbors algorithm displayed the highest accuracy rate of 97.83%, accompanied by 100% precision and 94.12% recall. Meanwhile, the Random Forest algorithm exhibited an accuracy rate of 93.48%, complemented by precision and recall values of 100% and 82.35%, respectively. The study's conclusions bear relevance for refining loan approval processes and fostering improved lending practices within financial institutions like CU KABARI.