Loss Estimation for Voluntary Motor Claims Based on Decision Tree Model

The objective of this research is to develop a model to predict own damage loss for voluntary motor claims. The dataset used in this research was collected from an insurance company for five years during 2014-2018. The decision tree based on the Chi-square Automatic Interaction Detector algorithm (CHAID) is selected to develop the loss estimation model. From the decision tree model, the data is classified into 43 groups. It is found that factors affecting the model are car brand, sum insured, responsibility, claim service type, number of third-party cars, cause of loss, extra services, and region. In addition, the loss estimated from the decision tree model for the test data set yields a percentage error of 4.01% compared to the actual loss for the claim data in 2019. Therefore, the decision tree model is feasible to predict loss of voluntary motor claims.