Research On Sales Forecasts Of Cars By BP Neural Network

The automotive industry is one of the four mainstay industries of our country, and it plays an important role in the national economy. Along with the slowdown of macroeconomic, automobile industry in China also begins to slow down. At the same time, the supply and demand of the automotive industry is imbalanced. Therefore, a reasonable sales forecast of the automotive market is crucial.BP neural network is a representative method of artificial intelligence. It has a good ability to predict nonlinear problems. It is selected as the main line to forecast monthly and annual car sales.The SARIMA model is used to forecast monthly car sales, and then BP neural network. It shows that BP neural network and SARIMA model have a good predictive accuracy. To further improve the prediction effect, BP neural network and SARIMA model are combined to establish a hybrid forecasting model. The results shows that the predictive capability of the hybrid forecasting model is better than a single forecasting model.For the forecasting problem of annual car sales, GDP, the per capita disposable income of urban residents and other six economic indicators are used for sales forecasting. The gray correlation analysis is employed to analyze the association between selected indicators and car sales. It is shown that the production of industrial raw material is most related with car sales.Then these indicators are standardized and are used as input for BP neural network. Because the redundancy of information is a problem for multidimensional data, principal component analysis is employed to reduce the data dimension. As a result, the MAPE of forecast fall from 2.34 percent to 1.71 percent, which shows the generalization ability of neural network has been enhanced.The forecasting of monthly and annual car sales shows that BP neural network is reasonable and effective in car sales forecasting problems.