A0974
Title: Introducing new rate factors and statistical learning to improve fire loss prediction accuracy
Authors: Ah-ram Lee - Ewha womans university (Korea, South) [presenting]
Abstract: Under the current insurance practice, the premium is charged based on a simple calculation rule using a limited number of rating factors in South Korea. While varying depending on the specific type of insurance, the rate-making procedure is not generally based on statistical models. Particularly in the case of fire insurance, the risk premium is calculated from a simple classification based on building type and structure. Furthermore, such risk classification is solely based on the frequency, and the modelling of the severity part is ignored. The insurance risk is modelled separately in terms of frequency and severity. The regression model and neural network approach are used to quantify the risk. Then, it is shown how to calculate the insurance premium based on these models.
