Research On Automobile Sales Prediction Model Based On Big Data

The production and operation activities of enterprises need accurate prediction of sales prediction.Accurate sales prediction can effectively reduce the waste of production resources and inventory accumulation.With the development of technology,the way of gathering information has gradually changed from offline word of mouth to online search engines.This change provides an opportunity for the use of online web search data and online reviews,making it possible for consumer behavior studying using web information.The web search data has been proved to contribute to automobile sales prediction.This paper introduces historical sales,Baidu index and sentiment data to predict car sales and see the difference on sales prediction of three different kind of data.According to the cyclical fluctuation characteristics of sales data,we adopt SARIMAX model to predict sales.Besides,considering the high dimensional characteristics of Baidu index,the convolution neural networks,which is good at feature extraction,is used to sales prediction.The result shows that both kinds of models work well,but the convolution neural network has better performance.What’s more,the hybrid data of Baidu index,historical sales and sentiment data performs best.The car sales prediction model based on hybrid data in this paper provides a new idea for the automobile sales prediction.This method not only contribute to automobile sales forecasting,but also can be applied to other industries.