The performance of one dimensional Naïve Bayes Classifier for Feature Selection in Predicting Prospective Car Insurance Buyers

Abstract

One of the products sold by insurance companies is car insurance. To offer this product, one of the techniques used by the company is cold calling. This method often decreases the sellers' mentalities because they face many rejections when offering insurance products. This problem can be reduced by classifying prospective buyers' data first. The data can be classified as customers with the potential to buy insurance and customers who have no potential to buy insurance. From the obtained data, there are certainly many features that support the classification process. However, not all features contributed to improving classification accuracy. Machine learning especially the method of feature selection helps to reduce dimensions and to improve classification accuracy. In this paper, we examine One-Dimensional Naïve Bayes Classifier (1-DBC) as a feature selection method that is applied to two classifier methods, i.e., Support Vector Machine and Logistic Regression. Our simulations show that the two classifiers can use fewer features to produce comparable accuracies in classifying prospective car insurance buyers.