臺灣博碩士論文加值系統

需求預測在供應鏈規劃中扮演著極為重要的角色，企業會在需求預測時使用聚合與分解的方式，以求減輕預測的負擔並增加準確度，然而要如何聚合、預測與分解，目前卻沒有足夠的參考依據。因此本研究探討聚合與分解策略在需求預測上的應用，希望了解如何能使需求預測效率與準確度獲得提昇。
本論文將以時間序列定義產品需求型態，分別針對平穩序列及非平穩序列，探討如何利用聚合與分解策略以提昇需求預測效能。在建立評估預測準確度的模型後，分析產品需求型態的各項指標對預測準確度的影響，其中平穩序列所使用的指標有：數列間的相關係數、期望值比例、標準差比例﹔非平穩序列的指標則有：數列間的成長趨勢比例、殘差項的相關係數、殘差項之標準差比例。
透過數學模型的分析以及實際資料驗證後，本研究得到一些使用聚合與分解策略的決策依據：
1. 對多筆水平型態的需求序列使用聚合策略時，只要不是所有數列的相關係數都等於1，就能使預測準確度改善。
2. 對多筆線性趨勢需求序列使用聚合策略時，只要不是所有數列的殘差項相關係數都等於1，就能使預測準確度改善。
3. 若針對兩筆水平需求序列使用聚合與分解策略，當相關係數越小、變異係數越接近時，預測準確度提昇程度越高。
4. 若針對兩筆線性趨勢需求序列使用聚合與分解策略，當兩數列間斜率越接近、殘差項之相關係數越大時，越適合使用聚合與分解策略。

Demand forecasting plays a very important role in supply chain management. Nowadays, supply chain managers try to reduce forecasting burden and get more accurate forecasts by aggregating and disaggregating product demands. But there aren’t sufficient rules about how to aggregate and disaggregate demands appropriately. This research explores the contribution of demand aggregation and disaggregation strategies to forecasting efficiency and accuracy.
This thesis defines demand patterns as time series models, and then studies stationary and non-stationary time series respectively. After developing the performance index to evaluate forecasting accuracy, we investigate the relationship between the characteristics of series and the performance of demand aggregation and disaggregation strategies. The characteristics of stationary series discussed are: coefficient of correlation, mean value, and standard deviation. The characteristics of non-stationary series discussed are: slope, coefficient of correlation of error terms, standard deviation of error term.
Some conclusions are obtained from the analysis:
1. Aggregation strategies applied to multiple stationary time series will improve demand forecasting accuracy except when every coefficient of correlation between two series equals to 1.
2. Aggregation strategies applied to multiple non-stationary time series will improve demand forecasting accuracy except when every coefficient of correlation between the error terms of two series equals to 1.
3. Aggregation and disaggregation strategies applied to two stationary time series will work better as the two series getting less correlated and the coefficients of variation of two series getting closer.
4. Aggregation and disaggregation strategies applied to two non-stationary time series will work better as the slopes of two series getting closer and the error terms of two series getting more correlated.

致謝……………………………………………Ⅰ
論文摘要………………………………………Ⅱ
Abstract………………………………………Ⅲ
目錄……………………………………………Ⅳ
表次……………………………………………Ⅵ
圖次……………………………………………Ⅷ
第一章 緒論……………………………………1
1.1研究動機……………………………………1
1.2研究目的……………………………………2
1.3研究方法……………………………………4
1.4論文架構……………………………………5
第二章 文獻探討………………………………6
2.1（s, S）存貨控制策略……………………6
2.2聚合與分解策略之相關文獻探討…………7
2.3聚合策略之相關文獻探討…………………9
第三章 平穩序列………………………………11
3.1模型架構與基本說明………………………11
3.1.1模型假設…………………………………11
3.1.2符號說明…………………………………13
3.1.3模型架構…………………………………14
3.2聚合與分解策略……………………………14
3.3聚合策略……………………………………30
第四章 非平穩序列……………………………35
4.1模型架構與基本說明………………………35
4.1.1模型假設…………………………………35
4.1.2符號說明…………………………………36
4.1.3模型架構…………………………………37
4.2聚合與分解策略……………………………38
4.3聚合策略……………………………………52
第五章 實證分析………………………………57
5.1平穩序列實證分析…………………………57
5.1.1資料簡介…………………………………57
5.1.2假設的驗證………………………………59
5.1.3應用聚合與分解策略進行預測…………60
5.1.4應用聚合策略進行預測…………………65
5.2非平穩序列實證分析………………………68
5.2.1資料簡介…………………………………68
5.2.2假設的驗證………………………………70
5.2.3應用聚合與分解策略進行預測…………71
5.2.4應用聚合策略進行預測…………………74
第六章 結論與建議……………………………78
6.1結論…………………………………………78
6.1.1關於平穩序列……………………………78
6.1.2關於非平穩序列…………………………79
6.2研究貢獻……………………………………80
6.3研究限制……………………………………81
6.4建議…………………………………………81
參考文獻 ………………………………………83

[1] 林聰明, 吳水丞, “指數平滑法之選擇與應用”, 華泰, 七十年十月
[2] 徐佳豪, “需求之聚合與預測策略研究”, 台灣大學工業工程學研究所碩士論文, 九十年六月
[3] 葉小蓁, “時間序列分析與應用”, 八十七年一月
[4] 鄭怡超, “需求之組合規劃策略研究”, 台灣大學工業工程學研究所碩士論文, 九十年六月
[5] George E. P. Box, Gwilym M. Jenkins, Time Series Analysis: Forecasting and Control, Prentice Hall, 1994
[6] Sunil Chopra, Peter Meindl, Supply Chain Management: Strategy, Planning, and Operation, Prentice Hall, 2001
[7] Charles W. Gross, Jeffrey E. Sohl, “Disaggregation Method to Expedite Product Line Forecasting”, Journal of forecasting, Vol. 9. 233-254, 1990
[8] Gerald Keller, Brian Warrack, Statistics for Management and Economics, Duxbury Press, 1997
[9] Edward A. S., David F. Pyke, Rein Peterson, Inventory Management and Production Planning and Scheduling, Wiley Europe, 1998
[10] David Simchi-Levi, Philip Kaminsky, Edith Simchi-Levi, Designing and Managing the Supply Chain: Concepts, Strategies, and Case Studies, McGraw-Hill, 2000