REGIONAL ANALYSIS OF CAR INSURANCE IN MONTENEGRO USING DEEP LEARNING METHODS

Abstract

This paper analyzes the regional distribution of the number of policies, total premiums and claims of 19486 legal entities as car insurance customers of one of the largest non-life insurance companies in Montenegro. Insureds are clustered using the k-means methods in four clusters: Poor-low risk, Middle-low risk, Wealthy-middle risk and Luxury-high risk. The clusters are described using the Decision Tree (DT) models. Then, an analysis of the distribution of these clusters by regions is made. The results show that in the southern region, compared to the other two, there are fewer insureds who have a small number of policies, low premiums and low claims (Poor-low risk cluster), while more are from the Middle-low risk cluster. Also, the cluster of insureds with high premiums and high claims is most prevalent in the northern region. It is shown that deep learning methods can be used efficiently for this kind of analyses.