Research On Prediction Of Several Kinds Of Automobile Sales Based On Fusion Baidu Index Multi-source Data

The prediction of automobile sales volume has always been a hot research problem.In recent years,due to the impact of the epidemic and other factors,China’s automobile sales volume has shown a downward trend as a whole.However,with the increasing awareness of national environmental protection,as well as the introduction of a number of policies to promote the development of new energy vehicle industry,China’s new energy vehicle market is becoming better and better.At the same time,as an emerging industry,its market competition pattern has not been fully formed.Therefore,predicting the sales volume of new energy vehicle and various brands of vehicles is not only convenient for policy makers to grasp the existing new energy vehicle market performance and future trends,but also helpful for automobile enterprises to reasonably plan resources and determine future development strategies.The overall monthly sales volume of new energy vehicles,BMW,Chery and BYD are selected as the research object in this paper.First of all,we collect text data and comment data from relevant automobile websites,and determine jointly the core keywords and expand the core keywords to build a key thesaurus based on the technical word selection method,supplemented by the direct word selection method and the scope word selection method.Then,crawling Baidu index data,through data cleaning,person correlation coefficient and time difference correlation analysis,the keywords with strong correlation with the corresponding car sales volume and leading time difference are selected.The results show that there is a significant correlation between Baidu Index and automobile sales volume,which can reflect consumers’ consumption intention.Next,we use the principal component analysis method to synthesize the keyword Baidu Index,and construct the keyword Baidu Index comprehensive index to predict automobile sales volume.Finally,the monthly sales volume of corresponding vehicles is predicted by using Least square regression,Random Forest,Gradient Boosting Regression,BP neural network and Stacking model,so as to compare the prediction effects of each model,and predict the monthly sales volume of corresponding vehicles in 2022.The results show that the Stacking model has better prediction effect than other single models,and its MAPE and RMSE have different degrees of reduction.Compared with the best single model,the MAPE of the corresponding vehicles dropped by 1.26%,3.84%,2.22% and 0.94%respectively.In the research objects,the accuracy of the model prediction is ranked from high to low as the overall sales volume of new energy vehicles,BYD,BMW and Chery.According to the prediction,China’s new energy vehicle market will continue to maintain rapid growth in 2022,and the sales volume of Chery and BYD will also show a growth trend.Due to the late layout in the new energy vehicle market,the sales volume of BMW will remain unchanged basically or decline.