- 學位論文
木製框式車身銷售數量之組合預測研究
A Study of Combined Forecast of Sales Volume of Truck Wooden Boxes
摘要
銷售預測的結果是許多重要生產決策的基礎，良好的預測結果能使整體的產銷規劃井然有序，成本得以控制，企業的獲利與長期的發展是可預期的；而糟糕的預測結果則可能導致不好的規劃，不是增加閒置成本與庫存成本，就是增加缺貨成本與流失客戶。 對木製框式車身製造業者來說，僅僅依據過往的銷售數據使用統計預測方法，再加上直覺的判斷，來預測未來的銷售數量，已無法因應多變的銷售環境。因此，本研究將以專業產製木製框式車身個案公司為對象，討論不同預測方法其預測狀況與適用性，使個案公司能從中選擇較優良的方法，來計算預測數值。 本研究在預測方法的選擇，除了以業界常用的移動平均法與指數平滑法外，另納入近來廣為預測研究使用的灰色理論。最後採用組合預測之三種預測模型，將移動平均法、指數平滑法、灰預測法所求得之解作組合預測，以探討組合預測模型是否適合木製框式車身的銷售數量預測。 由實證分析結果發現，以迴歸方式決定權數之組合預測模型不論是在訓練區間、績效測試區間、全區間，都有較佳的預測效果。因此，以迴歸方式決定權數之組合預測模型，為本研究最優良之銷售數量預測模型，適合且足可作為個案公司木製框式車身產品之銷售數量預測之用。
並列摘要
The result of sales forecast underpins vital production decisions, in which accurate forecast results in complete and orderly planning of overall production and sales operations, precise control over cost and predictable enterprise profit and its long-term development. Contrarily, inaccurate forecast can cause defective planning, either increasing idle and inventory cost, or enlarging stockout cost and losing customers. It has been difficult for truck wooden boxes manufacturers to respond effectively to volatile marketing environments by predicting future sales volume through only statistics methods based on past sales data as well as intuitive judgment. Therefore, this research, which focuses on professional manufacturers of truck wooden boxes, aims to discuss the forecast status and adaptability for variant forecast methods in hope of providing the manufacturer a more accurate method to calculate forecast results. Widely used in forecast research recently, Gray System Theory is also adopted as one of the forecast methods, besides the two widespread methods of Moving Average and Exponential Smoothing in this field. Based on the three prediction models, the combined forecast is conducted using the separate solutions of Moving Average, Exponential Smoothing and Gray Prediction to investigate whether the combined forecast model matches the sales volume prediction of truck wooden boxes. Empirical analyses reveal a combined forecast model, with weighting factor determined by recursive algorithms, offers better prediction in training scope, performance test scope and the whole scope, which justifies that the aforesaid model with weighting factor decided by recursive algorithms is the best one, and is suitable for sales volume prediction of truck wooden boxes.