AUTHOR=Manathunga Vajira , Zhu Danlei TITLE=Unearned premium risk and machine learning techniques JOURNAL=Frontiers in Applied Mathematics and Statistics VOLUME=8 YEAR=2022 URL=https://www.frontiersin.org/journals/applied-mathematics-and-statistics/articles/10.3389/fams.2022.1056529 DOI=10.3389/fams.2022.1056529 ISSN=2297-4687 ABSTRACT=

Insurance companies typically divide premiums into earned and unearned premiums. Unearned premium is the portion of premium that is allocated for the remaining period of a policy or premium that still needs to be earned. The unearned premium risk arises when an unearned premium is insufficient to cover future losses. Reserves allocated for the unearned premium risk are called premium deficiency reserves (PDRs). PDR received less attention from the actuarial community compared to other reserves such as reserves for reported but not fully settled (RBNS) claims, and incurred but not reported (IBNR) claims. Existing research on PDR mainly focused on utilizing statistical models. In this article, we apply machine learning models to calculate PDR. We use an extended warranty dataset, which comes under long-duration P & C insurance contracts to demonstrate our models. Using two statistical and two machine learning models, we show that machine learning models predict reserves more accurately than the traditional statistical model. Thus, this article encourages actuaries to consider machine learning models when calculating PDRs for the unearned premium risk.