Research on Risk Prediction Models Based on ARIMA and XGBoost

In this paper, the underwriting strategy of the insurance industry under extreme weather is studied, and the underwriting strategy is analyzed by ARIMA model, TOPSIS entropy weight method and XGBoost model. Firstly, the ARIMA model is used to predict the number of losses and deaths, and the sensitivity analysis of the per capita insurance cost is carried out to provide more accurate information to support the insurance industry to make more flexible underwriting strategies. We conducted validation in the United States and Africa to predict what insurers will cover over the next seven years. Secondly, considering the diversity of extreme weather in different regions, CGDAM-WRIR model combines underwriting policies with regional characteristics to maximize insurance benefits. Finally, TOPSIS entropy weight method and XGBoost model are used to score and predict buildings in order to provide reliable suggestions for the community. This paper provides theoretical support for the insurance industry to reasonably formulate underwriting strategies under extreme weather conditions, and emphasizes the importance of cooperation among government, insurance companies, communities and other relevant departments.