Research On The Prediction Of Claim Amount In Auto Insurance

With the further development of reform and opening up,the Car Parc is also increasing rapidly,which has also led to an increase in traffic accidents.As the most important one of insurance industry in the property insurance industry,the auto insurance is playing a critical role.The more traffic accidents,the more auto insurance.However,the auto insurance of increasing compensation,the decrease in profit margins and the increase in operating costs caused by traffic accidents have made profit margins less and less.At this time,we need to determine the insurance premium by predicting the amount of claims.The specific analysis of this thesis is as follows:1.Data Pre-processing.The original data is subjected to operations such as missing values,anomaly processing,and classification feature transformation.Perform analysis of descriptive statistical and influencing factor on the data,and analyze characteristic relation through bar charts,box plots,and arrangement charts;use random forest algorithm to analyze feature importance,and draw a histogram showing the importance of variables.2.The Tweedie regression model and zero-adjusted inverse Gaussian regression model of the claim amount were constructed by using six variables including policy type,gender,age,driving age,vehicle age and driving area.3.The machine learning algorithm model of claim amount was built.Normalized Mean Square Error(NMSE)was used as the evaluation standard to judge the merits of the model and build a random forest regression model,and the NMSE was 0.343.The support vector machine(SVM)regression model was constructed,and the NMSE was 0.336.After optimization by grid search and particle swarm methods,the NMSE was 0.301 and 0.273.The BP neural network model was built,and the NMSE was 0.309,0.269 after genetic algorithm optimization,and 0.263 after particle swarm optimization.In this thesis,Tweedie regression model,zero-adjusted inverse Gaussian regression model and machine learning model are used to predict the amount of auto insurance claims,and it is found that the effect of machine learning model is better than Tweedie regression model and zero-adjusted inverse Gaussian regression model.Taking the normalized mean square error(NMSE)as the evaluation standard,the BP neural network optimized based on particle swarm optimization is selected as the optimal model,which provides a certain reference value for the determination of the auto insurance rate.