Heterogeneous Multiple Classifiers Mengunakan C4.5, K-Nearest Neighbor dan Naïve Bayes untuk Menentukan Tingkat Pembaharuan Polis Asuransi Jiwa

Abstract

At a time when the insurance business is increasingly competitive, it requires insurance companies to have innovations in increasing the number of customers. With information from existing customer data, insurance companies can make decisions in implementing company strategies, including determining insurance customer decisions on the sustainability of life insurance policies. Data mining can form a pattern or create a trait of business behavior that is useful for decision making. In this research a Heterogeneous Multiple Classifiers prediction model was built using Majority Voting by combining C4.5, K-Nearest Neighbor and Naïve Bayes to determine the renewal rate of life insurance policies. The Heterogeneous Multiple Classifiers model that was built produced an accuracy value of 94.61%,  precision value of 95.20%,  recall value of 94.60% and an F-Measure value of 94.60%. The performance value generated by the Heterogeneous Multiple Classifiers based prediction model is higher than the performance value of the Single Classifier based prediction model. It is hoped that this method can increase the income of life insurance companies, for example by offering a promotional program for insurance policy renewal to customers who are predicted to extend or not to extend their insurance policies.