Pembobotan Atribut Pso Untuk Optimasi Svm Dalam Kasus Kelayakan Kredit Bank

Abstract

Credit success in a bank plays a role in maintaining the survival of a bank. Therefore it is very necessary to measure creditworthiness accurately to classify customers with good credit and bad credit. Based on these conditions the right data mining technique to use is classification. One of the data mining classification techniques is Naïve Bayes Classifier (NBC), but the accuracy is still less than the C4.5 and SVM algorithms. This final report describes the steps of research using the Particle Swarm Optimizatin (PSO) algorithm to weight attributes to increase the accuracy of SVM. This study uses data set public German Credit Data. The validation process uses tenfold-cross validation, while testing the model using confusion matrix and ROC curve. The results show SVM accuracy increased from 74.6% to 76.50% after combined with PSO.