行動裝置所涉及的產業及供應鏈十分廣泛，其出貨量足以代表資訊科技產業的整體性發展。環顧全球資通訊產業的發展，國際市場上的行動裝置出貨量備受期待，舉凡智慧型手機、平板媒體裝置、筆記型電腦及穿戴式裝置等，均在近幾年呈現爆發性的崛起，本研究分別採取出貨量最大宗及發展趨勢具代表性的智慧型手機及平板電腦作為研究對象，正因其影響力甚大，故有必要對行動裝置出貨量趨勢變化進行系統性的分析及預測。面臨如此龐大的商機，本研究旨於發展出有效的預測模式，除了高精確度外，必須容易使用，並符合成本效益，進而探討行動裝置市場的發展趨勢。其實證結果對於有預測出貨量需求的相關產業而言，可說明各模型運用於預測的適用性，是相當具有實務性價值的參考依據。 本研究以2010年至2015年之全球平板電腦及智慧型手機出貨量資料進行實證分析，採用時間序列分析法中的Decomposition模型、Holt-Winters模型、Moving Average模型及灰色理論中的GM(1,1)模型進行預測，並主要以2010至2014各20季的出貨量統計資料作為建模預測基礎(訓練期)，2015年各4季出貨量統計資料作為驗證預測模型成效(驗證期)。 實證結果顯示:1.在訓練期間，各模型皆發揮不錯的預測效果，整體又以應用於智慧型手機出貨量預測最佳，其模型預測能力可達良好及高精確度狀態。2.在驗證期間，受平板電腦出貨量下滑波動及智慧型手機出貨量成長減緩的影響，難免干擾各預測模型的準確度。3.其中又以平板電腦出貨量的2015年驗證期，受出貨量下滑波動的影響最為顯著，各單一模型預測準確度皆受干擾，卻經由組合預測後，大幅提升預測結果準確度並且比各單一模型預測效果更佳。 ;The industry and supply chain involved in mobile devices is very extensive .The shipments of mobile devices represent the overall development of information technology industry. Along with the development of the global information and communication industry, the shipments of mobile devices on the international market are expected to rise. In this research, the smart phones and tablets which represent the largest volume of shipments and the trend of development were taken as the research subjects. Given that their great influence, it is necessary to systematically analyze the trend of mobile devices shipments. Facing such a huge business opportunity, this study has developed effective forecasting models that explore the market trends of mobile devices. The aim of this study is to establish an excellent predictive model that must be easy to use, accuracy and cost-effective. For the related industries with forecasting demand, this study can be used as a reference with the practical value. The research methods in this study were the time series models, including Decomposition, Holt-Winters, Moving Average models and the grey theory of GM(1,1) method. In addition, the shipments of 20 quarters as training period from 2010 to 2014; the 4 quarters of 2015 as validation period. The results of study are as follows: 1. all the models have a good predictive effect during the training period. In particular, the model prediction ability can reach high precision state in smart phone shipments. 2. The decline in shipments of the tablet and the growth of smart phone slowed down that interfere with the accuracy of the forecast models in the validation period 3. Since the tablet shipments in the 2015 verification period is more significant decline with shipments fluctuations, this study via the forecasting combination model which can enhance the accuracy of the forecast results greatly. The predicted ratio of forecasting combination model is better than each predicted ratio of single model as well.
