Research On Regional Automobile Sales Forecasting Integrated Model

The Chinese automobile industry has always been an important part of the national economic growth.Its production and sales rank first in the world,and which has a profound impact on national life and social and technological progress.The increase in automobile sales directly affects people's production and life.However,in recent years,the growth of automobile sales has shown a trend of differentiation in time and space.How to effectively predict the growth trend of automobile sales in different regions and at different times is essential to assist local governments and automobile-related enterprises to formulate corresponding development strategies.Therefore,this paper studies the auto sales forecasting model with regional characteristics from a regional perspective,improves the applicability and accuracy of the model for regional forecasting,and allocates resources efficiently and reasonably.First,analyzing and studying the linear or nonlinear exponential function,power function,S-curve function,hyperbolic function characteristic trend between the 16 influencing factor data and car sales growth data in 30 provinces,municipalities and autonomous regions except Tibet,and according to the mixed kernel function to improve the prediction accuracy,the data is small sample,multi-attribute and other characteristics,The paper proposes to select Gaussian kernel function,Polynomial kernel function and Sigmoid kernel function to structure a mixed kernel function by linear weighting,and establish a support vector machine integrated model for regional automobile sales forecast.Then,the particle swarm optimization algorithm is used to optimize the parameter values of the regional auto sales forecasting integrated model,and finally it is determined as the PSO-SVM regional auto sales forecasting integrated model.Finally,Comparing and analyzing the prediction results of PSO-SVM forecasting integrated model and traditional support vector machine model and other models through the experimental simulation.The comparison model includes Gaussian kernel function SVM,Polynomial kernel function SVM,Sigmoid kernel function SVM,Gaussian kernel and Polynomial kernel mixed kernel function SVM,Gaussian kernel and Sigmoid kernel mixed kernel function SVM,Polynomial kernel and Sigmoid kernel mixed kernel function SVM,Multilayer Perceptron neural network and Long Short-Term Memory neural network.The experimental results show that the PSO-SVM forecasting integrated model can not only make up for the shortage of traditional support vector machines for regional auto sales forecasting,achieve the best accuracy of regional forecasting results,but also improve the applicability and accuracy of regional auto sales forecasting.It is effectively verified that the PSO-SVM regional auto sales forecasting integrated model is a model with regional characteristics.