Research On Short-Term Demand Forecasting And Inventory Allocating For Fresh Product

The purpose of this paper is to help solve the problems accuring in the circulation of fresh products,such as the low degree of organization,the long trading cycle and the low degree of informationization.The feasibility of the CPFR supply chain inventory management mode to solve the circulation dilemma of fresh commodities was discussed,and the technical support link in the process of "demand realization" was studied to improve the implementation effect by improving the accuracy of sales forecast and the rationality of inventory allocation,so as to promote the implementation of CPFR mode in the field of fresh commodities.Firstly,by combing fresh problems existing in the process of commodity circulation and analyzing the characteristics of the CPFR supply chain management mode of operation,the paper raised the possibility of improving fresh commodity circulation problems though the mode of CPFR,from the aspect of sourcing,quality control,transaction cost,loss of goods,inventory management,and so on.Then,combined with fresh characteristics,the importance of the accuracy of fresh commodity sales forecast and the flexibility of inventory allocation for the effective implementation of CPFR was pointed out.Secondly,in order to improve the reliability of the sales forecast results in guiding the inventory allocation process,the forecast accuracy was improved from two aspects of the characteristic engineering and the forecast model improvement.Based on micro-influencing factors of fresh sales,feature creation,screening and dimensionality reduction were carried out to obtain the feature matrix with high explanatory degree of target variables and low correlation between features.ARIMA model is used to describe the linear relationship contained in the sales volume time series.Furthermore,with the help of characteristic matrix,the NARX neural network model was used to mine the non-linear relationship contained in the ARIMA residual based on the residual compensation idea,so as to seek for higher prediction accuracy.Then,under the condition of secondary hub and spoke inventory allocation system,an integrated inventory allocation optimization model is established and an intelligent optimization algorithm is designed to solve the problem.Taking the sales forecast result as the initial target vertical allocation quantity,and considering the available inventory and allocation capacity limitation,the decision about the quantity of vertical and horizontal allocation for the target commodities in each FMC was made.According to the characteristics of the model,a three-layer code was designed to represent the feasible solution,and genetic algorithm was used to solve the model.Finally,based on the operation data of a fresh e-commerce company,the fresh product sales forecast scheme and inventory allocation strategy proposed in this paper were verified,and a series of problems that need to be solved before the company introduces the CPFR model were put forward from the perspective of forecast and inventory allocation.