Risk Modeling Journey - GLM and Impact Analysis

Generalized Linear Models (GLMs) are not a new topic. With increases in computing power and access to big data, actuaries have in fact been using GLMs in the insurance rating process for many years. Besides being well established, GLMs have straightforward interpretation which helps the communication with underwriters and the Product department. Although many theoretical works have been done regarding GLMs in the insurance business, it is also important for actuaries to explore this subject from a work experience perspective. In this context, this project aims to present a practical view of the GLM model-building process from start to finish. This work is applied to the Liberty Seguros motor insurance policy database and different risk models are built considering the automotive perils with higher exposure. The model evaluation step shows how well the model built aligns with historical data, which means that it is possible to verify the model's ability to predict the risk behavior for new datasets after model development. During the impact analysis, it is shown the suggested changes in the prices for specific variables and how it affects the overall price, which can help the company to improve the underwriting process and profitability.