在日趨競爭的環境中,企業為求生存無不竭盡所能,盼能研擬出最佳之決策以締造佳績。決策之制定必須掌握完善的資訊為基礎,才不至於流於盲目,推動企業成長的是銷售,唯有銷售不斷的成長,企業才能不斷的進步,預測具有展望未來之功能,其成敗往往左右企業的發展,建立一個有效之銷售預測系統,提高其預測績效,得以確實掌握市場脈動,減少生產過剩所造成之資金積壓的問題與缺貨所造成之訂單流失等問題,以提升企業的市場的競爭力作為目標。 本論文的目的在以類神經建構一預測能力佳之銷售預測模式。本研究以類神經網路為主體,加入網路修正因子α、β,作為誤差之修正項,加以變更網路權重,以期降低其各月份預測值之誤差,以國內某知名車用油品公司旗下之經銷商實際銷售數字來作為驗證,我們發現修正因子法與傳統類神經Re-training方法比較後,傳統Re-training法之預測績效沒有修正因子法為佳。
In the competing environment with each passing day, all enterprises try to exist as hard as they can, and expect to research and develop the best policy in order to create excellent achievement. To formulate the policy, we must have perfect information in hand basically. Thus it wouldn’t get blind. To promote the enterprise mature is selling. Only the selling grows up continually will the enterprise improve continuously. “Forecast” has the function of looking into future; success or failure would influence the enterprise development. To establish a efficient forecasting system of selling; to raise the forecast effect in order mater the market migration; to lessen the problems of asset overstock that cause from surplus of produce and the run off of orders that cause from shortage of goods; to raise the market competition of enterprise as a goal. The target of this thesis is to uses Neural Networks to set up a selling forecasting model that has well forecasting efficiency. This research uses Neural Networks as a subject, add α and βto be the revise item of error, and expect to lower the error of predictive value each month. Take the actual selling digit, the domestic branch agency of a famous lubricating oil corporation, to proof. After comparison of revise factor model and traditional Re-training method, we found that the forecasting efficiency of than traditional Re-training method.