Customer Churn Prediction Models For PT. XYZ Insurance

High persistence is a major business objective that wants to be achieved by insurance companies. To achieve high persistence, insurance companies need to be able to maintain their customers. Currently, the majority of insurance companies in Indonesia don't have an appropriate system in place to retain their customers. This paper presents an approach using data mining techniques to help such a company in dealing with customer churn. We explore some data mining techniques using real data from PT XYZ, an Indonesian based insurance company. The techniques implemented in this work are supervised learning methods, which are Decision Tree (DT), Naïve Bayes (NB), and Artificial Neural Network (ANN). It turns out that DT produces the best result with 91.3111% accuracy and 0.970 of AUC, followed by ANN and NB. With this level of accuracy, it is recommended that PT XYZ Insurance and Indonesian insurance company in general, may use the Decision Tree method for the customer churn prediction model.