Research On Prediction Of Non-life Insurance Claims Based On Random Forest

With the rapid development of artificial intelligence,it has brought various degrees of impact and impact to various industries and institutions,and its influence in the insurance field cannot be underestimated.The machine learning algorithms and ideas involved in artificial intelligence also subtly influence some traditional actuarial methods.At present,the application of machine learning technology in the actuarial field of insurance is one of the hot topics in the international actuarial theory.Due to the flexibility of machine learning technology in processing information,it makes it more adaptable to individual differences.This paper attempts to provide a new direction for the individual claims reserve of insurance by using the CART method and random forest method in machine learning.Such forecasting results take into account the differences among individuals and the insurance Pricing is also important.Firstly,the thesis combs the domestic and foreign researches related to non-life insurance reserves,and combines the international actuarial theory to study the hot spots and the current domestic research deficiencies in this field,so as to determine the overall research ideas and research methods of the paper.Then,some of the traditional actuarial methods of non-life insurance reserves were introduced and analyzed in detail.Then the classification and regression tree(CART)method and the random forest method in machine learning technology were carried out in terms of the advantages and disadvantages of algorithm steps and algorithms.Very detailed analysis.Finally,by using the relevant data information of the individual insurance claims,using the CART algorithm and the random forest algorithm in the machine learning algorithm to predict the number of claims,the prediction results of the number of claims based on individual information are obtained.By comparing the prediction accuracy of the CART algorithm and the random forest algorithm,we found that the accuracy of the random forest's prediction depends on the matching of the data volume and the number of trees;at the same time,the trees in the random forest Under the same number of conditions,for a small amount of data,random forests can improve the prediction accuracy of CART to a certain extent;but in the case of a large amount of data,the accuracy of random forest prediction is not as good as that of a single CART.The accuracy of forecasting is high.At this time,the number of trees in the random forest needs to be increased to increase the accuracy.