Spatiotemporal Visualization,Mining And Prediction Of Fmcg Logistics Data

At present,China’s fast-moving consumer goods industry has developed more mature: there are many kinds of products and fast circulation,but its unit value is low,and the profit is very thin.In order to reduce the cost,enterprises begin to work hard in logistics.Therefore,the accurate establishment of logistics volume prediction model is of great significance to guide the production and operation of enterprises and the vehicle scheduling of logistics companies Taking the pure water data of a certain brand as the research object,this paper analyzes the correlation of the logistics data,and analyzes the temporal and spatial evolution characteristics of the logistics shipment volume from the time dimension and spatial dimension.SARIMA model,LSTM model,SVR model and sarima-lstm combination model are respectively used to predict the pure water logistics shipment volume,and the prediction results are compared and analyzed.Firstly,in the aspect of pure water logistics data analysis,the correlation between the total annual logistics volume and the total resident population of each region is analyzed by Pearson correlation coefficient analysis method;in the aspect of time dimension,the change trend of national logistics volume is analyzed from three time scales of year,quarter and month,and the change of logistics volume is analyzed from the perspective of temporal and spatial evolution Finally,the periodicity of purified water logistics is obtained by wavelet analysis.Secondly,in the aspect of pure water logistics volume prediction,due to the nonlinear and linear characteristics of logistics volume data,this paper uses four models to predict the volume data(1)In the process of modeling with product season SARIMA model,firstly,the range of each parameter of product season model is determined by processing the stationarity of logistics shipment volume with difference and testing the white noise.Secondly,the optimal parameters are selected by AIC quasi measurement,and the residual error is tested to determine the prediction model.The MAPE of the prediction result is 8.2%,and the prediction error is still large,the main reason for which is the material shortage The non-linear characteristics of the data are obvious.(2)According to the non-linear characteristics of the data of logistics shipment volume,a prediction model based on LSTM neural network is constructed.The parameters of LSTM neural network are selected through experiments,and the logistics shipment volume is predicted.The MAPE of the prediction result reaches 5.88%,and the prediction effect is significantly improved compared with SARIMA model.(3)Because SVR has strong learning ability in processing time series data with small sample size,this paper establishes SVR model based on PSO optimization algorithm,and the MAPE of prediction result is 5.12%,which is better than SARIMA model and LSTM neural network.(4)In order to further improve the prediction accuracy,this paper designs the sarima-lstm combination model to predict the logistics shipment volume.The MAPE of the prediction result is only 3.52%,and the prediction accuracy of the combination model is higher than that of the single prediction model.