Research On Influencing Factors Of Passenger Vehicle Sales Based On Semi-parametric Model

As one of my country's important pillar industries,the automobile industry plays a pivotal role in the national economy.Automobiles not only expand the radius of people's trips,bring convenience and comfort to people's lives,but also influence the development of social productivity.After China's accession to WTO in 2001,the auto industry in the market has experienced a spurt of development,with the total production and sales of automobiles spanning from 3 million units in 2002 to 25.76 million units in 2019.The rapid development of the auto industry in recent years has prompted many auto companies to blindly expand their production.Blind production will not only cause waste of human and material resources,but also shake the whole national economy.To effectively solve the problem of auto overcapacity,this paper establishes a semi-parametric regression model to accurately predict auto sales by analyzing the factors influencing auto sales,which is of great significance to the healthy and sustainable development of China's auto industry.First,this paper analyzes the main factors affecting China's auto sales,and selects five major influencing indicators: consumer price index,consumer confidence index,steel production,average fuel price and private car ownership,and qualitatively shows that the five major influencing indicators are the causes of auto sales through Granger causality test.The grey correlation coefficients are calculated by grey correlation analysis to quantitatively demonstrate that the five major influencing indicators are more strongly correlated with auto sales,with steel production and average fuel price having the strongest correlation with auto sales.Secondly,this paper uses monthly data of auto sales from January 2010 to December2018 as well as five major impact indicators,conducts descriptive statistical analysis and normality test,predetermines two semi-parametric regression models by correlation coefficient matrix and co-curvilinearity matrix,determines linear and non-parametric terms of the models by solving linear regression models and non-linear regression models,and then performs parameter estimation and non-parametric estimation for the linear and non-parametric parts of the two semi-parametric regression models to obtain two semi-parametric regression models for predicting auto sales.Finally,data related to auto sales from January 2019 to July 2020(excluding February2020)are selected as the test set,and two semi-parametric regression models are used to make forecasts of auto sales separately,and by comparing the mean absolute error,mean square error,and goodness of fit of the two semi-parametric regression models,it is concluded that semi-parametric regression model 2(consumer price index and consumer confidence index are placed in the same non-parametric term)is a relatively good fit for auto sales.