Research On Demand Forecasting Model Of Arima-bp Neural Network Based On Cpfr

In retailing industry, CPFR (Collaborative Planning Forecasting and Replenishment) was consequently initiated to downsize the supply chain operation costs and increase customer satisfaction as well. CPFR is decided to improve internal and external cooperation efficiency, and lay a foundation for the full integration of supply chain. CPFR not only saved the supply chain operation costs, but increased the sales as well, and more important, led to supply chain integration by continuously improving the partner's relationship and trust level by extending cooperation scope from inventory management to demand management. Inter-enterprise collaborative demand forecasting is the core of CPFR implementation, it improve the accuracy of demand forecasting. And accurate demand forecasting can lower the production costs, transportation costs and inventory costs, can increase the sales, thereby improve the supply chain operation efficiency.The article studies forecasting model mainly based on Autoregressive Integrated Moving Average (ARIMA) and Back-Propagation Network (BP Network) of CPFR, on the basis of summing up the past, which takes needs as object including suppliers and more retailers over the supply chain, specific research contains:(1) Based on collaborative forecasting of CPFR, which achieves information sharing between the suppliers and retailers, and guarantees the accuracy of forecasting data, and realizes unification of prediction method between the two according to establishing project team, to the maximum extent, reduces the possibility of predicting anomalies.(2) The supply chain demand of the study, neural network as a forecasting tool, genetic algorithm as optimization tool, three of the basic theory and principle of the integration. After optimized the weight of neural network, the efficiency of learning greatly improved.(3) This article will be both organic:the use of ARIMA model sequences of linear prediction extraction of uncertainty, with the remaining residual characteristics of the nonlinear improved BP neural network model used information extraction, and establish ARIMA-BP forecasting model. By an example this article makes a forecasting result comparison between single ARIMA, BP neural network and the model, discusses the effectiveness of the ARIMA-BP.This article discusses the ARIMA-BP neural network forecasting model, and provides a new method for collaborative forecasting.