The Research For Second-hand Car Price Problem Based On Semi-parametric Model

With the vigorous development of China's economy,the rising of the people's living standard,cars are regarded as an indispensable means of transportation for business trips,travel and citizens' holiday trips,and with the change of people's consumption idea,diversified consumption,rational consumption,personality become mainstream,buying a used car also gradually become more and more people a new choice.However,problems such as the irregularity of the domestic second-hand car trading market and the imperfect evaluation system have puzzled many people.Therefore,the study of factors affecting the value of second-hand cars and the degree of influence of each factor on the value of second-hand cars is of great guiding significance to the establishment of second-hand car appraisal and evaluation system and the establishment of a healthy and standardized second-hand car market.In this paper,second-hand car transaction data in January 2018 on the Uxin Second-hand Car trading platform are selected to screen out the relevant data of German and American second-hand cars of the types of second-hand cars traded,and the training set and test set of the data set are selected in the ratio of 4:1.In this paper,the semi-varying coefficient model and semi-parameter partially additive model are used to analyze the price of German and American second-hand cars respectively,The scatter diagrams and fitting curves are used to summarize that there is complex relationship between prices and the influencing factors(linear and nonlinear relations as well as the influencing factors of the interaction between the relationship).The semi-varying coefficient model can not only reflect the linear and nonlinear relationship between variables,but also reflect the interaction between variables.On the study of the price of Germany,this paper proposes a semi-change coefficient model and a profile least squares estimation method to explore the price of new cars,displacement,transmission type,emission standard,mileage and usage time on used car prices.The analysis results found that the price,displacement and transmission types have a positive linear impact on used cars.Also,mileage,time and emission standards have complex nonlinear effects for second-hand cars.At the same time,the using time has an interactive impact with the mileage and emission standards,which together have an impact on the price of second-hand cars.Semi-parametric partially additive models include linear models,additive models and partially linear models,which have powerful modeling ability and flexible interpretation ability.Also,mileage,time and emission standards have complex nonlinear effects for second-hand cars.At the same time,the using time has an interactive impact with the mileage and emission standards,which together have an impact on the price of second-hand cars.When studying the price of American second-hand cars,this paper uses a half-parametric portion to increase the effect of six influencing factors on the price of used cars.It adopts backward proposed estimation calculation,and uses training samples to get US second-hand car prices and six Interpretation of the fitting regression relationship between variables.The analysis results found that the new cars price,displacement,emission standards,and transmission types have a positive linear impact on the US second-hand car,and the use time and mileage have a complex nonlinear impact on the used car.Finally,the model assessment and contrast are performed.By comparing the prediction accuracy of the two models,it is concluded that the prediction precision of the semi-coefficient model is better than that of the semi-parametric partially additive model in fitting the regression relationship between second-hand car prices and six explanatory variables.