Asia Pacific University Library catalogue

A thesis submitted in fulfillment of the requirement for the award of the degree of Master of Science in Data Science and Business Analytics (UCMF1808DSBA).

Every year billions of dollars are lost in the American auto insurance industry due to fraud, which forces insurance premium prices to go up annually. Although fraud detection solutions have been developed to fix the fraud detection problem, they all still face the same well­known problems of imbalanced data. Nonetheless, there is need for a centralized claims database to gather a holistic view of fraudulent characteristic behaviour. To tackle both of these issues, this research proposes a data pre-processing technique, particularly a fraud behaviour feature engineering approach, to improve the overall performance of prediction models. The behaviour being assessed is be based on the RFM model along with an additional behaviour analysis related to policy expiration. Furthermore, an ensemble feature selection and modelling is used to deal with the high dimensionalil:y problems that the feature engineering approach brings along with it, as well as the class imbalance problems. The proposed behaviour features along with the stratified bootstrap ensemble model show a very significant improvement in the overall model's performance compared against the published stat-of-the-art results. Precision increased by 57.3% (71% vs. 13.3%), Recall increased by 9.7% (90% vs. 80.3%) and the over model's performance from the F-measure improved by 56.2% (79% vs. 22.8%). This research work contributes to existing knowledge in the context that it helps to extend the scope of behaviour analysis used in the RFM model in analysing fraudulent behaviour with respect to auto insurance claims.