Statistical Modeling Of Returns's Distribution And Influencing Factors Of Internet Financial Products

In this thesis,Yu'e Bao,the largest and most representative one,is selected as the research object of statistical modeling to explore the characteristics of internet financial product.A hierarchical bayesian mixture distribution model is established to study the distribution characteristics of Yu'e Bao's return rate.In order to explain the distribution characteristics,a semi-parametric additive model is established to study the influencing factors of Yu'e Bao's return rate.The main research of this thesis is divided into the following two parts.The first part is to study the distribution of Yu'e Bao's return rate.Firstly,the mixture normal distribution with multiple peaks is selected to express the bimodal characteristic,which is found by kernel density estimation and histogram.In order to reduce the complexity and improve the accuracy of parameter estimation,a hierarchical bayesian mixture distribution model is established and can be seen in formula(2.6).The MCMC algorithm is used for bayesian estimation of parameters,and the results are shown in formula(2.14)to describe the specific form of bimodal distribution.Secondly,in order to verify the reliability of the model,the bayesian test is carried out.And the posterior chance ratio and bayesian factor of the test can prove that the data support the null hypothesis,that is,two means parameter estimated value of the marginal posterior distribution can fall into the two-mean 0.95 maximum posterior density confidence sets of the established distribution model,indicating that the model is reliable.Finally,based on the layered bayesian mixture model,the distribution characteristics of Yu'e Bao return rate are analyzed in depth.In order to explain the distribution characteristics,it's necessary to continue study the influencing factors of Yu'e Bao's return rate.The second part is to study the influencing factors of Yu'e Bao's return rate.Firstly,based on the above distribution characteristics,six variables are selected from the operation mechanism of Yu'e Bao and the market environment.A semi-parametric additive model is established after analyzing the linear and nonlinear relationships and interactions among variables.OLS estimation is used for the parameter part of the model,and smooth spline estimation is used for the non-parameter part,the model's expression is shown in formula(3.7).Secondly,in order to avoid the spurious regression which is caused by the instability variables,inspired by the idea of co-integration and error correction,a long-term equilibrium relationship among variables is discussed and a short-term relationship in formula(3.11)is adjusted.The reliability of the model is proved by residual test.On the one hand,from the inside of the model,based on the results of parameter estimation,the effects of various influencing factors are analyzed as follows: the net cash ratio,the Shanghai interbank offered rate and the Shanghai and Shenzhen index have linear effect to Yu'e Bao's return rate,the effect of Shanghai interbank offered rate is positive,and the other two are negative;the consumer price index has nonlinear effect of volatility;the narrow monetary supply and the USD/ CNY have a nonlinear interaction effect;the short-term fluctuation of Yu'e Bao's return rate comes from the current fluctuation of the respective variables,the pure random fluctuation in the current period and the error of the model in the previous period.On the other hand,from the outside of the model,this thesis makes an in-depth analysis of the factors influencing the return rate of Yu'e Bao and finds that each factor is driven by its own specific policies and market environment,which can well explain the reason why the return rate shows the characteristics of bimodal distribution.And also,the return rate of Yu'e Bao from January to March 2019 is predicted,and the relative error of the model is quite small,which indicates that the model has strong prediction ability.Since the sample range of modeling almost covers the whole development of Yu'e Bao,the model can be used to predict its future return rate for a long time.The two models can complement and verify each other.And,the hierarchical bayesian mixture distribution model can be further applied to other financial models based on distribution,and the semi-parametric additive model can provide reference for regulators and investors.This thesis not only enriches the quantitative studies of Yu'e Bao,but also can be used for reference in other relevant studies of internet finance,and has certain reference value for participants in various markets.