Font Size:
Multivariate Data Modelling for Car Insurance Sales Prediction
Last modified: 2020-02-13
Abstract
Selling car insurance is one of the important processes in an insurance company. The companies mostly need to call a potential customer many times to get the answer. It would be better to know the potential customers with high possibility to accept the offer. This information may be very helpful for the company to decide the call order of the customers. This study aims to predict the tendency of the customers in accepting the insurance offer. We have a data set of a private bank which has a call center and also tries to sell car insurances. This data set includes parameters like age, gender, living city, application channel, number of calls by the sales representative, number of days the process completed and so on. The output labels defined in the data set are sold and not sold. The number of not sold labels is more than the sold ones which makes the data set unbalanced. The main purpose of this study is first to analyze and prepare the data set to be used in a classification problem. We use some techniques to balance the data set such as ClassBalancer, Oversampling, Undersampling, SMOTE, and so on. Then, we apply several classifiers of different data mining techniques based on Bayesian, Regression, Rules, Trees, and Ensembling. In addition, a multivariate data partitioning and modelling technique called High Dimensional Model Representation (HDMR) based algorithms are also applied to the considered data set in order to generate a prediction model for the potential car insurance sales. Benchmarking results through several performance metrics are performed and some necessary comparisons are discussed in this study.