Optimasi Penilaian Kredit Nasabah dengan Metode Expectation-Maximization-Naïve Bayes

Abstract

Customer credit data has not been optimally utilized to identify patterns that can be used to predict the eligibility of new credit applicants. One of the main challenges is the absence of class labels in customer credit data, which hinders the classification process. This study aims to develop a data mining model that combines the EM (Expectation-Maximization) clustering method and Naïve Bayes classification to predict the eligibility of new credit applicants. The EM clustering method is used to assign class labels to unlabeled data, enabling the classification process with Naïve Bayes. From a total of 540 customer credit data points analyzed, 142 were classified into cluster 0 (non-eligible credit applicants) and 398 into cluster 1 (eligible credit applicants). The results indicate that the combined method achieved an average accuracy rate of 99.24% and an average error rate of 0.76%. Using the WEKA software, this study concludes that the combination of EM clustering and Naïve Bayes classification is an effective approach for predicting the eligibility of new credit applicants.