成功導入六個標準差不僅可改善公司產品之品質,並降低生產成本,能為企業帶來競爭優勢。但有鑒於許多企業於推行六標準差(Six Sigma)時面臨許多挫折與困難,本研究運用多屬性方法及貝氏網路新法建構六標準差導入風險之預測模式。首先,本研究以文獻為基礎,彙整企業導入六標準差之重要因素,藉由模糊德爾菲法(Fuzzy Delphi Method, FDM),經量化分析及篩選的程序獲得具關鍵性的影響因子,作為導入六標準差的評估準則。接著,採用詮釋結構模式法(Interpretive Structural Modeling, ISM)分析導入六標準差因素間的關連性,有系統地以圖形化方式呈現變數間複雜的因果關係。進而,為使不同背景的專家評估值能更客觀呈現,本研究以層級分析法(Analytic Hierarchy Process, AHP)結合灰關聯分析法(Grey Relational Analysis, GRA)求得各專家權重。再經由模糊積分(Fuzzy Integral)整合各專家針對各因素之狀態機率與因素之間相互影響之條件機率之評估值,以建構貝氏網路方法(Bayesian Network, BN)導入六標準差之風險評估模式,希望所建構之模式可以提供欲導入或已導入六標準差而成效不彰企業之參考。所評估之企業只需將該企業導入六個標準差之目前的各因素狀態機率值填入本評估模式中,即可預測該企業存在之可能風險值。由多屬性方法及貝氏網路之敏感度分析皆可得知最重要的兩項關鍵因素為「高階主管的支持與參與」及「了解與滿足客戶的需求」,也就是說,企業若需導入六標準差可以先由此兩項關鍵因素先著手。在實際案例中,個案公司透過本研究所提出的模式評估了解到若目前推動六標準差其高度風險的機率為32.4%、中度風險為28.1%、低度風險為39.4%。
A success introduction of Six Sigma does not only improve the product quality but also reduce production costs and improve enterprises’ competitive advantages. Still, there is amounts of enterprises face many setbacks and difficulties wile promoting Six Sigma, the study uses multi-attribute methods and the Bayesian Network to construct and introduce Six Sigma to the risk prediction models. To step out, this study takes literatures as foundation, compile those important factors that lead enterprises’ introduction of Six Sigma, with Fuzzy Delphi Method, we have obtained the key factors as the assessment of introducing Six Sigma through procedures of quantitative analysis and filtering. Through Fuzzy Integral, we have integrated assessment values of experts on state probability of various factors and conditional probabilities which are interactively affected by factors, in order to construct a risk assessment model import of introducing Six Sigma through Bayesian Network, we hope the constructed model to offer references for those enterprises which have not been introduced Six Sigma yet and those have but ineffective. The assessed enterprises only need provide this assessment model all probability values of factors and status about introducing Six Sigma, then the possible risk value to that enterprise can be predicted.