專家權重法的DEA權重設限之研究

A Research on the Weight Restriction of DEA Incorporating Expert Weighting Method

邱木霖(Muh-Lin Chiou)

指導教授 ： 魏乃捷 ； 薄喬萍

義守大學/管理學院/工業管理學系/博士(2020年)

摘要

傳統上，資料包絡分析法(DEA)是利用數學的線性規劃方式，找出一組能讓受評單位(DMU)效率評估極大化之權重配置。至於各變項權重之區間範圍，也以DMU既有投入與產出數據取得。而此時所得到的數據資料，即相當於統計抽樣的結果，並非統計母體的真實狀態。
為了解決「DEA模式普遍具有變項權重為"0"」的不合理問題，以及能夠更吻合真實狀態的權重範圍評估；本文不但結合專家權重法(EWM)與保證區域(AR)模式，同時加入與「專家訪談」相關元素，包括：「專家諮詢、專家問卷、回饋式專家意見」的整體性作業方式。如此所產生可預期影響，至少包括：(1)遵循EWM的理論，以凸顯出各變項之間的關聯重要度；(2)區間估計的寬度會收緊(寬度愈小，準確度愈準)；(3)平均值的變化也不會太大。
如此新發展出的DEA模式，本文稱之為「分段式權重比例區間估計」- tEWM-AR模式。這樣的評估方法，不但具有「操作方便、直接接觸專家」的特性與好處。同時，也將是另一種近似統計母體真實狀態的評估方法。
研究結果發現，tEWM-AR模式的主要成果有三：(1)可以解決「變項權重為"0"」的不合理問題；(2)讓「兩兩權重比例區間估計」具非負下限的特性；(3)能夠吻合真實狀態的權重範圍評估。

關鍵字

資料包絡分析法 ； 效率評估 ； 專家權重法 ； tEWM-AR模式

並列摘要

Traditionally, Data Envelopment Analysis (DEA) uses mathematical linear programming to find a set of weights that can maximize the efficiency evaluation of Decision Making Unit (DMU). As for the range of each variable weight, it is also obtained from the existing input and output data of the DMU. However, the data obtained at this time, which is equivalent to the Statistical Sampling results, is not the true state of the Statistical Population.
In order to solve the unreasonable problem that the "DEA mode generally has the weight of variable is 0" and the evaluation of the weight range that can more closely match the real state. This article not only combines the Experts Weight Method (EWM) and the Assurance Region (AR) model, but also adds a set of holistic work methods related to "Expert Interviews"; elements include: "Expert Consultation, Expert Questionnaire, Feedback Expert Opinion." The predictable impact of this includes at least: (1) Following the EWM theory to highlight the connotation of the importance of the association between variables; (2) The width of the interval estimate will be tightened (the smaller the width, the more accurate the accuracy); (3) The average value will not change too much.
This newly developed DEA model is called the Segmented Weighted Proportional Interval Estimation - tEWM-AR Model. This evaluation method not only has the characteristics and benefits of "easy operation and direct contact with the Experts," but also will be another evaluation method that approximates the true state of the statistical population.
The research results show that tEWM-AR Model has three important results: (1) It can solve the unreasonable problem of "the weight of variable is 0." (2) Let the "Interval Estimation of Pairwise Weight Ratio" have the characteristics of non-negative lower limit. (3) Evaluation of the weight range that can match the real state. In other words, the tEWM-AR Model can provide decision makers with a new, objective, and unbiased model for efficiency assessment.

並列關鍵字

Data Envelopment Analysis ； Efficiency Evaluation ； Expert Weighted Method ； tEWM-AR Model

參考文獻

1.Chinese part

1.Ching-Kuo Wei, Rong-Kwei Li, Liang-chih Chen (2011); The Study on Optimal Weight of Data Envelopment Analysis: A Case Study of Taiwan Medical Center. Management & System, 18 (3), 505-527.

2.Education Dictionary (December 2000); Zheng Yongjing, Delphi Survey, http://terms.naer.edu.tw/detail/1313712/ (Web check time: 2020/04/22).

3.Song Wenjuan (2001), A Kind of" Equal Quality "Research Method-Application of Dephi Method in the Field of Medical Management Research, "Journal of Medical Management", Volume 2, Issue 2, pages 11-19.

4.Yan Qingyou and Tao Jie (2015); efficiency evaluation of China's biomass power generation industry. Operations research and management, 24 (1).