主流記憶體之二十年價格模式研究與驗證;Pricing Model of Two Decades of the Mainstream Memory Research and Verification

NCUIR > School of Management at National Central University > Executive Master of Industrial Economics > Electronic Thesis & Dissertation > Item 987654321/54882

Please use this identifier to cite or link to this item: http://ir.lib.ncu.edu.tw/handle/987654321/54882

Title:	主流記憶體之二十年價格模式研究與驗證;Pricing Model of Two Decades of the Mainstream Memory Research and Verification
Authors:	吳惠婷;Wu,Hui-Ting
Contributors:	產業經濟研究所碩士在職專班
Keywords:	記憶體;DRAM;NAND FLASH;NOR FLASH;平均單位位元價格;Memory;DRAM;NAND FLASH;NOR FLASH;Multiple;ASP
Date:	2012-08-21
Issue Date:	2012-09-11 19:09:01 (UTC+8)
Publisher:	國立中央大學
Abstract:	本研究目的為探討全球主要三種記憶體市場的價格變化影響因素。經由收集Gartner(高德納諮詢公司)之2000年-2010年11年間的DRAM ( Dynamic RandomAccess Memory) & NAND FLASH & NOR FLASH之全球年帄均價格以及相關出貨數量為研究依據，研究期間從1991年-2010年，共橫跨20年。本研究收集20 年資料並加以區分為10 年資料與20 年資料對照迴歸估計分析結果是否一致以及利用均方根誤差RMSE判斷二組迴歸分析估計式的準確度：(1) DRAM 最大宗應用量的個人電腦全球出貨量與全球DRAM 市場出貨量，以簡單迴歸來模擬DRAM 價格，並比較估計準確率；本研究實驗結果發現：1991 年-2010 年(20 年)資料估計出的DRAM帄均單位位元價格其RMSE 誤差值皆比1991年-2000 年(10 年)資料的RMSE 值要大，尤其以全球DRAM 市場出貨量為自變數時，RMSE結果差異最大。(2)而在綜合兩種本研究主觀認為最主要影響的變數後，進行二組資料複迴歸分析，其預測值與實際DRAM 帄均單位位元價格的RMSE值卻呈現20 年資料估計準確度比10 年資料估計準確度高出10 倍，與簡單迴歸分析估計RMSE 值結果方向不一致 (3)套用此迴歸模式繼續研究NAND FLASH帄均顆粒價格估計，對NAND FLASH 出貨量、NOR FLASH 出貨量、DRAM 出貨量進行複迴歸分析，得RMSE 可縮小至1.68，是本研究所有模型中均方根誤差最小，準確度最高的預測模型。This research studies on the major factors of market price of 3 popular worldwidememories. We estimate the price of the pasted 20 years by using the data comes fromGartner Dataquest and Market Intelligence & Consulting Institute ( MIC).We separate the data into 10 years(1991-2000) and 20 years(1991-2010), to checkif the 20 years data’s regression results and Root Mean Square Error (RMSE) is betterthan the 10 years’. The procedures of this research are listed as follows: (1) Using theDRAM’s major application – worldwide PC shipment quantity or worldwide DRAMshipment quantity as the base, estimate DRAM price by simple regression, we find theRMSE generated by 20 years data is worse than 10 years data, especially the case ofworldwide DRAM shipment quantity. (2) Using the two above varieties with 20 yearsdata in multiple regression; we fine that the 20 years’ RMSE(4.69) is better (small)than the simple regression’s RMSE(6.75~11.57). (3) We also study the influence onprice of NAND FLASH average selling price (ASP)/unit of the changes in NANDFLASH worldwide shipment quantity or NOR FLASH worldwide shipment quantityor DRAM worldwide shipment quantity: our results indicate that the accuracy in termsof RMSE is the highest among the empirical models in this study; RMSE is narroweddone to 1.68.
Appears in Collections:	[Executive Master of Industrial Economics] Electronic Thesis & Dissertation

Files in This Item:

File	Description	Size	Format
index.html		0Kb	HTML	958	View/Open

社群 sharing

Loading...