| Summary: | 碩士 === 國立中正大學 === 資訊管理學系 === 91 === Demand forecasting has been one uncontrollable issue among industry activities. Main reason lied in the over diverse of the product sale within a period, which makes so-called Forecasting Error remaining high. High forecasting error will severely affect whole supply chain activity of the industry. However, responding to the short circle of terminal products and the fast responding to customers, which causing compressed stock and enlarged demand of factories among the supply chain, excess of components of products and shortage and loss of material are resulted in the whole supply chain. Therefore, Bullwhip Effect is nothing more emergent issue of supply chain management.
This study collected related literature in the past, dealt with bullwhip effect by setting an integrative model, discussed further about effects of several main key causes to reflect real bullwhip effect on upper and lower layers among supply chain, so that the result can be provided as references for bullwhip solution. On the other hand, bullwhip effect on each member among supply chain can be references for thinking strategies as well.
This study tries to provide a novel thinking pattern and demand forecasting method to fix the problem of information exchange and accuracy, further to enhance accuracy of sale forecasting and lower bullwhip effect so that the efficacy of the whole supply chain will be lifted. On the other hand, by collecting, integrating, analyzing, and suggesting accurate and instantaneous information through demand forecasting system, industries can reach their goal to adjust their future sale forecasting, improve accuracy of forecasting, provide the most suitable net requirements for better efficacy, improve customers’ satisfaction, and lower cost and waste.
This study has confirmed that ”neural network based on Kalman filter” has faster styptic effect than traditional back-propagation neural network on demand forecasting, that it is not easily affected by error of variable setting, that it forecasts future demand more precisely, especially when input and output variables are more complicated nonlinear imply, which causing more diversity of efficacy between two models. Therefore, it is even more suitable to use this model for demand forecasting under the supply chain circumstance.
|
|---|
