Comparison Of Pso-Based Naive Bayes And Naive Bayes Algorithm In Determining The Feasibility Of Bumdes Credit

Abstract

In analyzing a credit, sometimes an analysis performs an inaccurate analysis so that some customers are less able to make credit installment payments, resulting in less smooth or even bad loans. From these problems, researchers conducted a credit analysis using computerized techniques using RapidMiner software in data processing. The right data processing technique to use is classification. One method of data mining classification is the Naive Bayes algorithm. Researchers use weighting by implementing Particle Swarm Optimization (PSO) for attribute selection to improve the accuracy of Naive Bayes. After testing with two models namely Naive Bayes algorithm and Naive Bayes based on PSO, the results obtained are for the Naive Bayes algorithm with an accuracy value of 93.24%, while the Naive Bayes algorithm based on particle swarm optimization models produces a higher accuracy value of 98.16% compared to the Naive Bayes algorithm model. From these results, the difference between the two models is 4.92%. Then for the results of using the ROC curve for both models, for the Naive Bayes algorithm, the AUC value is 0.939 with an Excellent Classification diagnostic level, and for the Naive Bayes algorithm model based on particle swarm optimization, the AUC value is 0.977 with an Excellent Classification diagnostic level. From the evaluation of the ROC curve, it is seen that the Naive Bayes model based on particle swarm optimization is higher when compared to the Naive Bayes algorithm. From the results of the AUC, the difference between the two models is 0.038.