| The vigorous development of e-commerce promoted the birth of return freight insurance,which quickly entered the public’s field of vision.As a new type of Internet insurance product,return freight insurance has solved the problem of consumer return disputes to a certain extent,but in the process of use,it has not brought the expected benefits to insurance companies.In this regard,this paper introduces the Bayesian network method into the pricing of return freight insurance from the perspective of online shopping consumers and based on the expected loss pricing method,and puts forward a pricing model of return freight insurance based on Bayesian network to realize the reasonable pricing of premium,which provides a new idea for formulating the pricing scheme of return freight insurance and has certain reference significance.This paper mainly uses Bayesian network method to establish the pricing model of return freight insurance,and verifies the effectiveness of the model.Firstly,this paper summarizes the factors affecting consumers’ return based on the literature,obtains the sample data required for the research through the questionnaire,and makes a descriptive statistical analysis.Secondly,build a return freight insurance pricing model based on the expected loss theory,and transform the return freight insurance pricing model into a key parameter that calculates the probability of consumer return.Then,divide the samples according to the commodity categories purchased by consumers,and divide them into three groups of samples: clothes,hats,shoes and accessories,digital electronics and beauty and skin care,and preprocess the data.Then a variety of Bayesian network structure learning methods are used to establish the Bayesian network structure,select the optimal Bayesian network model to predict the probability of consumer return,analyze the relationship between various influencing factors and consumer return,and verify the results of the model.The results show that the accuracy of Bayesian network prediction model under clothing,hat,shoes and accessories category is 74.32%,that under digital electronics category is 75.64%,and that under beauty and skin care category is 76.56%.The accuracy of the model is higher.Compared with other machine learning methods,the effect of Bayesian network model is better.This proves the effectiveness of the Bayesian network model established in this paper.After the return probability of consumers is obtained by Bayesian network method,the insurance rate of return freight insurance can be determined.Finally,based on the results of the above Bayesian network model,this paper uses Bayesian network reasoning to obtain the posterior probability of consumer return.For consumers with historical behavior data,it is more appropriate to use Bayesian network causal reasoning to calculate the posterior probability of return,while for consumers without historical behavior data(new online shopping users),it is more appropriate to use Bayesian network diagnostic reasoning to calculate the posterior probability of return.In addition,when applying Bayesian network to causal reasoning,this paper further excavates and analyzes the influence of different factors on consumers’ return probability under different commodity categories. |
