Finance: Theory and Practice | 2026
Authors: Filimonova I.V.; Komarova A.V.; Samatova A.P.
DOI: 10.26794/2587-5671-2026-30-2-108-120
Journal: Finance: Theory and Practice
Year: 2026
Publisher: Financial University under The Government of Russian Federation
Document Type: Article
Open Access: All Open Access; Gold Open Access
Cited by: 0
The purpose of the study is to develop and apply a model for predicting potential losses under a CTP policy using machine learning techniques. the relevance of the topic is due to the importance of the role played by the insurance market, and, in particular, transport insurance in the development of the Russian economy. This relevance is supported by the high loss ratio for this type of insurance, as well as the need to optimize market conditions. The object of this research is the Russian insurance sector. The subject of the study is machine learning techniques that allow predicting CTP losses based on input parameters characterizing data about the insured and their vehicle. The paper compares the effectiveness of ensemble machine learning methods with the traditional generalized linear method of predicting compulsory motor insurance losses. The study divides CTP losses into traditional and direct categories. It has been shown that the effectiveness of applying boosting machine learning models for forecasting is higher than using Random Forest and GLM. Factors that significantly affect the frequency and severity of insured events include: the number of minor accidents on a vehicle, the number of drivers, and the CBM coefficient. Other factors include the minimum length of service for a driver in a policy, the power of a vehicle, and the type of location where the vehicle is used. It has been concluded that the growth of the projected loss is positively influenced by the number of accidents, the number of drivers in the policy, low seniority and low age of drivers. The influence of the gender composition of drivers is highlighted: an increase in the number of female drivers leads to a decrease in insurance risks under the CTP policy. © Filimonova I.V., Komarova A.V., Samatova A.P., 2026.
CTP; loss forecasting; machine learning; risks; transport insurance