公司與顧客最直接且頻繁的關係,可認定為彼此之間的產品買賣關係,若沒有訂單,公司就無法獲取利潤,維持公司之存續就會出問題。因此訂單的獲得,是一切商業公司生存的基礎。接獲訂單後,各別產業才能依其生產系統,加以運作生產,在生產線的排程中,若是生產機台與人員無法協調,將會造成企業資源的浪費,因此需要訂單之預測以解決此問題。本研究使用實例公司之歷史資料,以移動平均法、指數平滑法、類神經網路和模糊類神經網路等四種方法比較其訂單預測的準確度。由實驗顯示,模糊類神經網路之準確度,優於其他三種方法,較適合做為訂單預測之工具,並且將此訂單預測工具應用於生產現場,以改善現場生產排程,提高生產效率與活化生產線。
The relation between an enterprise and its customers is orders of goods. No benefit will be gotten if there is no order. In addition, forecast of the number of orders is very important for an enterprise, because the number of orders affects production utility. A sudden large number of orders causes the paucity of materials, parts, and human powers. On the other hand, lack of orders causes the waste of enterprise’s resources. Using case data of a real company, in study applies four approaches, including Moving Average, Exponential Smoothing, Artificial Neural Network, and Fuzzy Neural Network, to forecast the company’s orders of goods. The results indicate that Fuzzy Neural Network approach has the best forecast results.