標題: | 結合資料採礦技術與貝氏信賴網路估計台灣積體電路設計公司之績效 Combining data-mining techniques with Bayesian belief network to estimate corporate performances of Taiwanese IC design companies |
作者: | 張家豪 王志軒 Chang, Chia-Hao Wang, Chih-Hsuan 工業工程與管理系所 |
關鍵字: | 積體電路設計產業;平衡計分卡;貝氏信賴網路;資料採礦;績效診斷;IC design industry;balanced scorecards;Bayesian belief network;data-mining;business diagnosis |
公開日期: | 2016 |
摘要: | 台灣的積體電路設計產業見證了該產業的過往興衰,台灣從最多曾高達400多間到2016年為止已經減少近1/3,且持續縮減當中;因此產業正面臨巨大的轉變,而公司也想藉由企業績效診斷從激烈的市場中存活下來,且從中取得更多的獲利,從各大軟體公司積極切入商業智慧的領域,可知其中龐大的商機。本研究以平衡計分卡為架構來衡量台灣的積體電路設計公司在財務、學習與成長、顧客及內部流程構面的表現,除此之外還利用資料包絡分析法來衡量營運效率並且納入內部流程構面,而在此四大構面之下總共高達24項的變數,因此找出會影響每股盈餘及股東權益報酬率的關鍵變數為優先重要的課題,本研究結合統計與資料採礦的方法進行兩階段的變數篩選,在分類器的部份比較了天真貝氏法、分類決策樹、貝氏信賴網路三者在分類績效上的表現,且利用分類決策樹來找出造成公司績效差異的關鍵變數及規則;在迴歸的部分則採用迴歸決策樹、多元適應性雲型迴歸、支撐向量迴歸三種方法來預測非線性結構的資料,研究結果發現支撐向量迴歸的表現最佳,且進一步利用多元適應性雲型迴歸找出的重要變數來進行改善後的績效模擬。此外也測試在不同調節變數(公司資本規模、主力產品強度)下之關鍵變數有何不同,最後並利用貝氏信賴網路找出重要變數之間的相互影響以便進行績效模擬,並提出公司績效管理的意涵。 Taiwanese integrated circuit (IC) design industries have witnessed the fluctuating rise and fall. In the past, the number of Taiwanese fabless design companies was up to 400. In 2016, the number has been reduced to nearly one third and continuing to shrink since the industry is facing a big change. Meanwhile, many software vendors are spending their efforts on business intelligence to help firms keep more competitive and earn more profit. In order to help Taiwanese IC design industries find key performance indicators (KPIs), this study combines the balanced scorecard (BSC) with the data envelopment analysis (DEA) to set 24 predictors. In particular, a two-phase approach is adopted to conduct feature selection. In performance benchmarking, class and regression tree (CART), Naïve Bayes Classifier (NBC), and Bayesian Belief Network (BBN) are compared. In performance regression, CART, support vector regression (SVR), and multivariate adaptive regression splines (MARS) are used to build nonlinear regression models. Furthermore, with consideration of moderating variables (i.e. capital size and product intensity), performance simulation is conducted by using SVR and BBN, respectively. |
URI: | http://etd.lib.nctu.edu.tw/cdrfb3/record/nctu/#GT070353344 http://hdl.handle.net/11536/138510 |
Appears in Collections: | Thesis |