Research On The Intensity Of Auto Insurance Claims Based On Machine Learning

Since the reform and opening up,China’s economy has been developing rapidly,people’s living standards have been continuously improved,basically reaching the well-off level,the number of car ownership is also increasing year by year,which leads to the motor vehicle insurance in the whole type of insurance products in the market share showing a growing momentum.Among them,the intensity and frequency of claim are the two decisive factors to measure the auto insurance premium,and the prediction research of these two factors is the top priority of actuaries.For a long time in the past,the generalized linear model was usually used to study these two factors.This model has certain validity in the research at that time.At that time,the research of auto insurance rate determination limited the explanatory variables to static factors such as people and cars.With the rapid development of data science and the gradual popularization of the Internet of Vehicles system,the acquisition of dynamic factors is more convenient,and the amount and dimension of available data have been significantly improved.However,the traditional method requires the assumption of data distribution before the establishment of the model,which is very difficult for the practical application of the large amount of data at present.Therefore,the feasibility of the traditional method is getting worse and worse,and it is difficult to make accurate prediction,thus affecting the determination of the final premium.At the same time,machine learning algorithms in the Internet industry have been widely concerned by actuaries for their excellent performance in efficiently and accurately processing large-dimensional data.This kind of algorithm solves the problem that the distribution of data needs to be assumed in advance in the traditional method,and can directly explore the law contained in a large number of dynamic data and apply it to the prediction research,which provides a practical method to solve the problem of auto insurance pricing in the practical application.To sum up,this paper introduces the decision tree algorithm and neural network algorithm in machine learning into the study of auto insurance rate determination.Based on the traditional static factor and the dynamic factor of the Internet of vehicles,it makes a prediction study on the claim intensity under the condition of loss occurrence,and makes a comparative analysis with the traditional generalized linear model.It is concluded that the prediction effect of the two machine learning algorithms is still better than that of the traditional model without considering the prior distribution of data.This shows that decision tree and neural network algorithm have broad application prospects in the field of auto insurance rate determination.Among them,the BP neural network model has the best fitting effect,which lays a model foundation for the study of auto insurance rate determination.After improvement,auto insurance premium pricing will be based on the indicators of different policyholders,design evaluation rules,comprehensive scoring,charging different premiums.This also helps the policyholder to regulate their own behavior,cultivate good driving habits,reduce the occurrence of accidents,promote the continuous development of the auto insurance industry,and promote the construction of a harmonious society.