<![CDATA[Today, classification performance has become increasingly important for credit risk assessment for loss control and revenue maximization. Therefore, a classification method is required that can accurately and efficiently measure the credit risk level of prospective borrowers as the key to the credit approval process. This study contributes to the development of feature selection methods with SI algorithms that use binary representation, namely feature selection using PSO algorithms with binary representation or Binary Particle Swarm Optimization (BPSO) applied to credit risk classification, with classification evaluation using kNN classification method. The application of feature selection is done to eliminate excessive features, thus reducing the number of features, improving the accuracy of the model, and reducing running time. The test results showed that KNN's best accuracy of 76.40%, can be improved by bpso-based selection feature with better accuracy of 88.70%, with an accuracy improvement of 13.35%. This test showed that bpso-based selection feature technique successfully improved the accuracy of KNN classification on credit risk classification.]]>
Copyrights © 2021
