Typhoon Loss Prediction Based On Gradient Boosting And Application Research Of Index Insurance Pricing

China has a vast geographical scope,and the natural disasters that occur every year are quite frequent.The safety and property of the whole society is threatened by the frequent occurrence of catastrophe.The influence of Typhoon disasters in the coastal areas of China mainly appears as the coefficient of winds,heavy r ain and storm surge,and the geological disaster and ecological disaster caused by the typhoon-rainstorm disaster chain,etc.According to the path of typhoon,wind,rain and other weather monitoring and forecast results,how to complete the typhoon disaster and risk assessment work,consequently improving the pertinence and efficiency of the deploy and strategy of preventing and mitigating the typhoon disaster,and mitigation of the pertinence and efficiency of deployment,and strategy,is a worthing subject.Therefore,it is of great significance to predict and analyze the typhoon loss accurately and effectively.This paper first describes the basic principle of Gradient Boosting algorithm and its advantage and feasibility of application in the typhoon los s prediction;Secondly,based on the dangerous hazard-formative factors and the vulnerability of hazard-affected bodies,this paper analysis the origin of typhoon disaster;Using the Gradient Boosting algorithm to forecast the typhoon loss,and analysed the prediction effect,it is proposed that a typhoon disaster database should be established and actively use big data and modern science and technology;Next,we explored the application in the premium rating of typhoon index insurance,applying Gradient Boosting algorithm to financial typhoon index insurance.We design an index insurance policy which is divided into two triggers and 6 hierarchies,input the triggers into the GB prediction model,using the level of 3% loss ratio to calculate the insurance amount.According to the frequency of typhoon as claim frequency to calculate pure insurance cost,provides a certain reference to the design and research of the financial typhoon index insurance.Finally,according to the empirical results,the recommendation of establishing a multi-party loss compensation mechanism based on insurance is given.In this paper,we synthetically uses relevant theory knowledge of economics,machine learning,statistics,actuarial science.In the research method,we combined machine learning and actuarial science,normative analysis and empirical.Based on Gradient Boosting algorithm in machine learning,we make an empirical analysis of the typhoon loss prediction in Guangdong province,prove the feasibility of the algorithm on the typhoon loss prediction from two aspects of theory and practice;Through the establishment of the loss prediction model,the application of the insurance rate of typhoon index is analyzed and discussed,combi ned machine learning and insurance actualization.This article apply Gradient Boosting algorithm in the typhoon loss prediction and typhoon index insurance pricing,the forecasting error rate remained at 1% to 15%.We have obtained the conclusion that this algorithm can be used to predict the typhoon loss and reduce the basis risk of typhoon index insurance,and have better application value in the typhoon disaster management.