本研究主要針對物流中心三種飲料類商品的出貨量作預測分析，以探討不同資料型態適用 的指數平滑技術，並找出各種商品需求量的影響因素，建立其多元迴歸分析預測模式，最後整合指 數平滑法與迴歸分析模式建立混合預測模式。本研究所採用的指數平滑法包含一次指數平滑法與二 次指數平滑法，並將三種商品之資料型態分成上升型、下降型與上下跳動型等三種型態，分別求出 指數平滑法、迴歸分析模式及混合法的預測值，並比較其差異。研究結果發現指數平滑法中二次指 數平滑法較適用於上述三種資料型態之商品預測，迴歸分析法在整體表現上則優於指數平滑法，而 混合法的整合模式則有最好的預測績效。

　　　　　This paper utilizes exponential smoothing method, regression method, and hygrid method that intergrating above two ones to forecast three types of drink product in a distribution center. Three types of product data including uptrend, downtrend, and fluctuation are employed to test the adaptability of single and double exponential smooth method. Regression method is also used to forecast these three types of data and comparing the forecasting results. The results show that double exponential smooth method is superior to single exponential smooth method, regression method is better than exponential smooth methods, and the hybrid method can achieve the best forecasting performance.