資料包絡分析法(Data Envelopment Analysis,DEA)方法可用來處理多項產出及多項投入指標的效率評估難題,無需先預設產出、投入指標權重值,所以,DEA在績效評估上常被應用至不同領域做績效評估,但DEA仍存在一些問題,例如(1)DEA在績效評估上有鑑別度不高的問題,(2)多個有效率受評單位(Decision Making Units,DMUs)產生多組權重解,(3)目標式只考慮到絕對效率觀點,忽略相對效率觀點。針對第一個問題,本研究以正規化權重向量模型來提高DEA的鑑別度;針對第二問題,本研究提出共同權重模式來改良多組有效率DMUs產生多組權重解的問題;針對第三個問題,本研究提出最小化排名模式的新排序模式來說明。 最後,本研究舉了二個實際的例子,來說明並比較原始DEA方法的評估結果與本研究所提出方法的評估結果。
Data Envelopment Analysis (DEA) is a tool of performance assessment manipulating multiple inputs and outputs without giving subjective weights of each input and output in advance. Although DEA has been applied in many fields to evaluate performance, the method still has following problems. i.e., (1) Result in to many efficient decision making units (DMUs) , i.e., low discrimination. (2) Obtain many sets of weight for DMUs. (3) Consider absolute efficiency instead of relative efficiency. This study adopts weight normalization approach to treat the problem (1) to increase the discrimination of DEA. To deal with problem (2) this study present a common weight model to find a set of weight for all DMUs with the maximum score. A rank minimization model is developed to assess the DMUs to eliminate the problem (3) mentioned above. Practical examples are also presented to illustrate the differences between the DEA and the proposed method.