| Summary: | 博士 === 國立臺灣科技大學 === 工業管理系 === 96 === Forecasting is an important part of supply chain management. The forecasting accuracy will influence the material planning and the scheduling; even affect the replenishment plan of the retailer of the supply chain downstream. In this study, we focus on two issues which are the collaborative forecasting between enterprises and the demand forecasting. The collaborative forecasting concept, shares the information and risk between the supply chain partners, combines both advantage the seller and buyer. It no longer only focuses on one side of supply chain, but contains the coordination, planning, execution and monitoring among all members until collaboratively producing the final forecasting with of both agreement. The final forecasting could drive the replenishment plan successfully.
Demand forecasting emphasized on how to promote forecasting accuracy of the products. We could obtain a more suitable demand plan and a better collaborative forecasting due to the higher degree of accuracy. If we promote the accuracy of collaborative forecasting and demand forecasting, we can monitor the forecasting accuracy and avoid the shortage or too many stocks in the whole supply chain. To improve the accuracy of the forecasting, we consider three forecasting problems which include collaborative forecasting, multi-product forecasting and combining forecasting under the demand forecasting.
First, we applied Six Sigma methodology and proposed a continuous improvement model on different phases of collaborative planning, forecasting and replenishment (CPFR). A real case is used to demonstrate how to improve the performance of collaborative forecasts.
Second, in a multi-product framework, the traditional estimation methods could not get the satisfied results. We have conducted research using the hybrid genetic algorithm (GA) for an efficient parameter estimation method for multi-product forecasting.
Finally, we developed a demand forecasting methodology that combines market and shipment forecasts. We used the LCD monitor sales data to test and verify our methods. In the past studies, the linear ways were usually used to estimate the parameters of combining forecasts. Fuzzy neural network (FNN) is a nonlinear model and often used to find the best combination in past references. The applications of FNN to combining forecasting problems are extremely few. We developed an integrated fuzzy neural network model to find the best combining forecasting and compare with other traditional methods such as k method, adaptive set of weights and linear composite. Results show that the proposed model using integrated FNN can gain the superior forecasting efficiency and performance in the whole.
|
|---|
