年度101
等級
書名運用SVR與Bass模型於台灣主機板與筆記型電腦預測分析
部分章節台灣是世界排名第一的主機板與筆記型電腦生產國家, 其在2011的佔有率分別為80.4% 與93.6%.而其未來之預測發展對高階主管而言是非常重要. 本研究目的即是提供有效預測模式給主機板,筆記型電腦製造商與相關產業作為未來擴產與投資之重要參考. 在本研究中, 我們運用支持向量回歸與貝氏擴散模式, 分析1998年至2012年台灣主機板與筆記型電腦季出貨之資料, 運用格子點參數搜尋,非線性最小平分法, 基因演算法與粒子群最佳化,分別找出支持向量回歸與貝氏擴散模式之最佳參數. 而運用平均絕對值百分比誤差(Mean Absolute Percentage Error, 簡稱MAPE)進行預測效益評估.本研究結果分析,支持向量回歸之平均絕對值百分比誤差低於貝氏擴散模式,可運用在主機板與筆記型電腦市場預測分析。
Taiwan is the world’s leading motherboard (MB) and notebook (NB) manufacturer, boasting a 2011 global market share of 80.4% and 93.6%, respectively. It is highly crucial for executives to predict future trends from within an environment of uncertainty. The aim of this study is to provide an efficient forecasting model to serve as a key reference for MB and NB manufacturers looking to expand or invest. We propose the following 2 forecasting models based on MB and NB quarterly shipment data from 1998-2012: (a) support vector regression (SVR) using a grid search method for the estimation of three parameters; and (b) Bass diffusion models (BDMs) using non-linear least square (NLS), genetic algorithm (GA), and particle swarm optimization (PSO) methods for parameter optimization. We also evaluate the forecasting accuracy by actual mean absolute percentage error (MAPE).The obtained MAPE values indicate that the proposed SVR model outperforms the BDMs using NLS, GA, and PSO for fitting and forecasting based on MAPE, and is therefore recommended for MB and NB market forecasting analysis.
出版社國立臺灣科技大學管理研究所
全部作者蕭裕耀
出版日期2013-06-01
所屬計畫案運用SVR與Bass模型於台灣主機板與筆記型電腦預測分析
備註專書