訂單需求預測-以時間序列分析研究

An Empirical Study on Order Forecasting : A Time-Series Analysis Approach Toward Demand

劉代華(Tai-Hua Liu)

指導教授 ： 陳飛龍

國立清華大學/工學院/工業工程與工程管理學系碩士在職專班/碩士(2008年)

摘要

摘 要
企業的經營目的在獲得利潤，企業必須做規劃來了解市場環境變遷的趨勢，掌握現在與未來的機會並避開威脅，才能獲得所期望的利潤。然而需求規劃是企業內部所有規劃的起源，所以預測是規劃活動的基礎，企業必須清楚市場需求才能有效展開各生產計畫，而企業的重大的決策也必須仰賴預測來做為參考依據。
企業進行各類預測分析時，主要分成定性及定量兩個方向，在進行定量分析時，主要是蒐集過去的歷史資料分析其發展趨勢，然而需求預測之歷史資料是複雜的，如何能發展出較快速、簡易、準確、不易受產業景氣循環影響的預測模式一直是企業努力的目標，為此本研究以時間序列法中的灰色理論預測、簡單移動平均法、加權移動平均法及雙指數平滑預測模式來建構預測模型，透過這四種方法的預測模式建構，分析比較找到最為合適的模型方法，並找出各種方法在訂單需求預測的限制，然後以預測之結果透過監控方式，調整訂單需求的方法。
本研究以一個案公司之歷史訂單資料為案例，以此來建構灰色理論模型、簡單移動平均法、加權移動平均法及雙指數平滑預測等四種模式，並使用絕對平均誤差及平均絕對誤差百分比來評定模式誤差，所得到之結果顯示雙指數進行平滑法最佳，其結果最為接近實際值，加權移動平均法次之，灰色理論預測第三，最後為簡單移動平均法。關於訂單需求之變化，本研究透過誤差管制圖及監控訊號來監控評定預測模式，然而對於預測模式仍有許多方法，未來研究方向將可朝向各種不同預測方法運用。

關鍵字

簡單移動平均 ； 加權移動平均 ； 灰色系統理論 ； 雙重指數平滑

並列摘要

ABSTRACT
The objective for an enterprise is to sustain profit by realizing marketing trend and grasping any business opportunities. To achieve this purpose, business planning plays an essential role. However, demand planning is the root for all internal plans and forecasting is the base for all the planning activities within the company. A good demand planning provide for a base for effective execution of business tasks. And it is needless to say that forecasting is the key reference for any important decisions made.
The forecasting methods can be classified into qualitative and quantitative approaches. While executing the quantitative analysis, a company is collecting historical data and analyzing the market trend. However, the historical data are usually very complicated. How to develop an easy, fast, and precise forecasting model which is not so sensitive to the rapid environmental changes has become an important target for enterprises. In this research, four time series models including Grey System Theory, Simple Moving Averages, Weighted Moving Averages and Double Exponential Smoothing will be analyzed and compared to find the most optimal model. Constraints of these four models in demand forecasting will also be studied..
An empirical study by a case company from her historical data was tested through the four methods, Grey System Theory, Simple Moving Averages, Weighted Moving Averages and Double Exponential Smoothing. The forecasting error was evaluated by the Mean Absolute Error (MAE) and Mean Absolute Percentage Error (MAPE). It is found that the Double Exponential Smoothing method has the best performance. Weighted Moving Averages is second, Grey System Theory is the third and the last is Simple Moving Averages. The study has also monitored and evaluated the forecasting models by error control chart and signal tracking. However, there may still exist other forecasting models that can be applied for the similar purpose and therefore it deserves further studies in the future.

並列關鍵字

Simple Moving Averages ； Weighted Moving Averages ； Grey System Theory ； Double Exponential Smoothing

參考文獻

許哲強(2002)，台灣區域電力負載預測分析系統之建立與應用研究，成功大學資源工程研究所博士論文。

Donlebell & Krasner, O. J.(1977), Selecting Environmental Forecasting from Business Planning Requirements, Academy of Management Review, 19, 373-383

Hamilton , James D. (1994), “Time Series Analysis”, Princeton University Press

Hartshorne, Robin (1997). Algebraic Geometry. Springer-Verlag

Martino, J. (1983). Technological forecasting for decision making. New York: Elsevier Science Publishing Company