建立DRAM供需預測模型

Abstract

本文以歷史資料，分析DRAM在供給、需求和平均價格、美國GDP等時間序列資料，進行實證分析，透過實證模型去了DRAM價格的變動的原因，本文透過了充足率、產業結構以及總體經濟的因素和誤差調整項目去架構出DRAM價格走勢的模型。 進一步的，由於科技產業需求端由過往PC為主轉變為Mobile Devices以及Cloud Computing的應用，使得DRAM的application更為複雜化，而這也對充足率以及平均銷售單價都有重大的影響。本文第三、第四章增強了歷史的模型，試著導入更多變異數。嘗試驗證可行性以後，首度將該驗證後的methodology帶到2015的預測上。 最後，本人僅以對DRAM產業的了解，對於後續研究者提出建議，並且分析現有的競爭態勢以及產品分析，做出對於未來產業結構變化的預測。

The thesis is based on the empirical analysis of chronological data of DRAM's supply, demand and average price as well as the U.S. GDP. Through the empirical model, it seeks to understand the reasons behind the DRAM price changes. The thesis constructs a DRAM price trend model based on sufficiency ratio, industry structure, macroeconomics factors and items for deviation adjustments. Moreover, as the IT industry demand transitions from PC-dominated applications to be driven by Mobile Devices and Cloud Computing, it is making DRAM's application even more complicated. These factors also have significant changes to the sufficiency ratio and average selling price. In Chapters 3 and 4 of this thesis, the historic model is strengthened, attempted by introducing more variances. After the feasibility has been proven, this is the first time that the proven methodology is incorporated into the 2015 forecast. Lastly, some recommendations are provided for the future researchers, based on my personal understanding of the DRAM industry. It also includes the analysis of the current competitive landscape and product analysis to arrive at a projection of the future changes of the industry structure.