供應商評選模式之建立，對於企業營運及提升競爭力，實具有相當重要之影
響地位；本研究之研究目的，為建立可用於實務上供應商評估之模式，藉此提升
企業競爭力，並將學術模式應用於解決企業實務問題。本研究將進行文獻探討，
以找出較具代表性之供應商評估準則，進而建立供應商評估問卷，將供應商分為A
類(表現優良廠商)、B 類(表現普通廠商)及C 類(表現較差廠商)，並以Likert 與Fuzzy
量表分別統計其準則分數。其次，以結構方程模式(Structural Equation Modeling,
SEM)，對於所挑選之評估準則與抽取樣本進行驗證性因素分析(Confirmatory
Factor Analysis, CFA)，觀察其樣本與因素結構配適度是否良好；再針對分析結果
進行變數篩選或資料轉換，以建立具同質性之評估準則模式，及具代表性的評估
資料。最後，將SEM 驗證後之資料，以支援向量機(Support Vector Machine, SVM)
進行分類模型建構；並且在經由模型訓練與測試後，與多變量統計中之複鑑別分
析(Multiple Discriminate Analysis, MDA)進行分類準確度之比較；以確認SVM 分類
模型在供應商評估模式議題上具有較高的效能。

Building the model of supplier selection is a critical role to improve
competitiveness and business process. The purpose of this study was to build a model of supplier selection to better encourage organizational capability and competitiveness and apply the model to solve practical problems. After literature reviewing, present study chooses the critical criteria for supplier evaluation and then develops the questionnaire
which differentiates A (excellent firms), B (common firms), and C (disappointed firms) from suppliers to evaluate by participants. The score was measured by Likert and Fuzzy scale. To understand the fitness between data and the measurement model, Confirmatory Factor Analysis (CFA) of LISREL statistical software package was calculated. The data and variables were transformed and divided to establish the model of homogeneous evaluative criteria according to the result. Finally, the data which confirmed by SEM were carried out to build the model by Support Vector Machine (SVM). After the model training and test, the results were compared with that from the Multiple Discriminate Analysis (MDA) of multi-variance statistics for classification precision, and confirmed that SVM classification model was more efficient in predicting the model of supplier
evaluation.