On the Cost-efficiency of Automobile Insurance Fraud Detection Methods: A Meta-analysis

Abstract

We suggest an improved cost savings calculation method in case of automobile insurance fraud detection. Moreover, we compare the cost saving ability of 77 fraud detection methods using heat maps and compare the cost saving ability of traditional statistics-based methods with that of machine-learning based methods, based on Romanian insurance data. The findings of this study clearly show that only a small percentage of the currently available automobile insurance fraud detection methods can be applied in a truly lucrative manner, and in general, machine-learning detection methods prove to be less cost-effective than the traditional econometric-statistic methods.