The Research On Enhancing Risk Classification Ratemaking In General Insurance With Data Mining

Radically speaking, Generalized Linear Models (GLMs) belong to the models of risk classification ratemaking. They are important methods in general insurance pricing. Today, GLMs are widely recognized as the industry standard method for pricing private passenger auto, other personal lines and small commercial lines insurance.All models of risk classification ratemaking are based on the loss data collection and risk classification, thus missing value, validity of risk identification and classification in the operation of insurers would influence the veracity of the models. Additionally, it is very difficult for the above models to evaluate the influence of risk factors, and to forecast the future losses. The author illuminated that the shortcoming of risk classification ratemaking models also existed in GLMs inevitably according to his empirical analysis.Data Mining is a process of finding various models, summarizing information, and exporting values from dataset. By recurring to Data Mining, the analysts can use a series of tools to analyze the data, get the feedback, and review the raw information in the new angle of view. There are five steps in Data Mining: Sampling, Exploring, Modifying, Modeling, and Assessing. Under the environment of SAS?, the author discussed the steps mentioned above thoroughly, and explained their meanings (of them) in terms of actuarial science.Also, this paper used SAS? Enterprise Miner? to give a series of demonstration. According to the algorithm of Regression, Decision Tree, and Neural Nets, the author rebuilt six ratemaking models, meanwhile, provided some explanations on how to improve each model. We compared the models from their Reorganization, Distinguish, Veracity, Stability, and Explanation. It also analyzed the intrinsic reason of discrepancies existed in these models.Data Mining can greatly enhance traditional ratemaking methods in general insurance, help the actuaries to achieve some important aims in ratemaking automatically, adjust the structures of rates, build a rational system of individual risk ratemaking. It will also help insurers identifying the insureds with high risk, thus reduce their loss ratios.