Statistical Learning for Predictive Modeling of Auto Insurance Claims

Abstract

The auto insurance branch is considered the principal and the first activity in an insurance company in terms of the claims paid and the premiums collected. For that, the insurers deal with a dilemma. They must price their automobile insurance products at their fair value, and at the same time, they must take profit from this pricing. Thus, pricing is a central issue for insurance companies, especially auto insurance. In the same way, the reinsurance companies care about the pricing of the claims transferred by the insurance companies for the purpose of making a profit. In this context, this article aims to model the costs and frequency of claims associated with the automobile insurance and reinsurance contracts, notably the pricing of the automobile liability guaranty, with a view of helping the actuary to get better pricing of reinsurance treaties, particularly "excess of loss" treaties using a new method. The approach to this work consists of the use of machine learning algorithms as an alternative to conventional methods. The statistical learning models are trained and tested on a large data-set that contains the explanatory variables. The results of this study demonstrate the advantage of applying statistical learning in the pricing of auto insurance and reinsurance contracts based on the predicted cost and the predicted frequency of the insured’ claims. The new mechanism presented in this paper is vital for reinsurance companies to get accurate charge pricing.
Copyright © 2022 Praise Worthy Prize - All rights reserved.