Market Demand Forecasting Considering User Behavior

The scale of cross-border e-commerce transactions in China continues to grow,accounting for an increase in the proportion of total imports and exports.The cross-border e-commerce has the market capture capacity and logistics of cross-border e-commerce companies due to the physical distance between the country and the country.Timeliness cannot be compared with that in a single domestic market.Especially in the case of selecting a maritime logistics channel,it is particularly difficult to grasp the demand for market demand.Therefore,it is particularly important to accurately predict the sales volume of the target market.This paper takes the commodity forecasting in the supply chain backup scenario of cross-border e-commerce enterprises as the research object,and combines the user behavior analysis in behavioral finance,including user value classification,user network behavior analysis,user commodity review analysis,search index,historical commodity sales data and commodity attribute data to forecast and analyze the market demand of commodities in the next month.After ARIMA,EMA,GBM and NN models are used to predict,and then Stacking model fusion is used to fuse the prediction results of multiple models in order to improve the accuracy of the overall prediction.The GBM model compares the prediction accuracy before and after adding user behavior data with that before adding user behavior data.The results show that time series models such as ARIMA and EM A are better than those which only consider the dimension of historical sales.The accuracy of time series models such as GBM is worse than that of multi-factor prediction models such as GBM.The prediction accuracy of GBM is further improved after considering user behavior data,which proves that more data and more information can bring better results.Finally,the model fusion results show that the accuracy of the fusion model is higher than that of each single model,and the greater the time span,the higher the accuracy of the fusion model.The fusion model is 22%higher than the benchmark model,27%higher than the time series model ARIMA,8%higher than the gradient hoist GBM and 39%higher than the neural network under the demand of 30 days.Finally,based on the economic bulk ordering model,the procurement strategies of different commodity categories are analyzed to realize the method of satisfying the market demand and saving the procurement cost and improving the economic benefit of the enterprise.