New Energy Vehicle Sales Forecast Based On LSTM Neural Network And Baidu Index

As one of the four pillar industries,China’s automobile industry is expanding its industrial scale,carbon emissions are increasing,and its energy dependence is far beyond the warning line.Under the dual-carbon target,it is inevitable to use the promotion and application of new energy vehicles to reduce the dependence on oil resources and realize the establishment of an automobile power.As one of the strategic emerging industries in China,the new energy vehicle industry has developed rapidly,leading the world in the industrial scale,and has a promising prospect.With the further promotion of network technology and the improvement of people’s material living standards,customer access to product information is no longer the word of mouth,but under the era of the Internet engine search,the transformation of information access makes the network information for both consumers and producers are effective research resources.Considering the availability of new energy vehicle sales data,this paper is based on the idea of data mining,to explore the era of byd new energy vehicles sales and baidu search index for the relationship between the sales forecast,build baidu index and LSTM neural network model combination model,based on the basis of data simulation fitting,to byd new energy vehicles for example of new energy vehicle sales data for short-term monthly sales forecast.Use China Association of Automobile Manufacturers to obtain the historical sales data of BYD new energy vehicles;use the highest search engine in China to collect the relevant keyword data of BYD new energy vehicles,and then conduct correlation analysis and time difference correlation analysis to select the keywords with the highest correlation and leading relationship for modeling research.After the collection of sales data and keyword data,ARIMA model,single-feature LSTM neural network model and combination model were established respectively based on the historical sales data,and linear regression equation and multi-feature LSTM neural network model were established by using Baidu index.The final prediction results showed that,The classical ARIMA model predicts a mean absolute percentage error MAPE of 10.99%,The MAPE of the single-feature LSTM neural network was 8.92%,The predictions have all achieved good results;Simple baidu index regression equation is more suitable for several short-term predictions;The MAPE for the combined model of ARIMA and LSTM was of 8%,The combination of the models effectively improves the accuracy of the prediction ground;The MAPE of the multi-feature LSTM model based on the Baidu index was6.64%,Is the best one of all the models in this article,However,the data workload is large and difficult,Daily predictions are available with ARIMA or single-feature LSTM models,It can also achieve good results.This paper uses Baidu Index and LSTM neural network for the sales forecast research of new energy vehicles of a single brand,rather than sticking to the overall sales forecast of new energy vehicles,refines the research objects,and improves the new ideas for the research in the field of new energy vehicle forecast.And the experimental results of this paper show that the above model can be achieved well in the sales forecast of a single brand of new energy vehicles.