非同质和多流程数据包络分析研究

发布时间：2017-12-31 09:41

本文关键词：非同质和多流程数据包络分析研究　出处：《中国科学技术大学》2016年博士论文　论文类型：学位论文

【摘要】：如何公平评价决策单元(Decision Making Units,DMUs)相对于同行的效率,一直是近几十年来研究的一个焦点。三十多年前诞生的数据包络分析(Data Envelopment Analysis,DEA),是一种有效的评价多输入和多输出决策单元的相对效率的工具,广泛应用于DMUs的评价和改进相对效率。许多研究者一直致力于利用DEA进行研究,成千上万的杂志文章和书籍面世。最近研究的一个特定领域是利用网络DEA测量DMUs的效率。两阶段DEA是网络DEA研究的一个重要的子集,众多的研究有不同形式的两阶段的分析。其中一种是封闭的串行流程,第一阶段所有输出作为中间变量进人第二阶段,成为唯一输入,没有任何元素离开此系统。研究的问题是如何将DMUs多阶段的效率拆分成单个阶段的效率值。一项重要的研究成果是利用博弈论和DEA理论结合去确定DMUs综合效率,进而确定单个阶段的平均效率。随后的重要科研成果之一是利用单个阶段效率几何平均和算术平均方法去确定DMUs总体效率。近年来另一个重要的研究方向是非同质DMUs效率,非同质现象有多种表现形式。例如：DMUs不同的输出组合和不同的输入组合。非同质DMUs综合效率第一步骤是DMUs通过内部结构划分为一套互斥的子决策单元,子决策单元应用和决策单元相同输入和输出指标,数据规模按分开的比例缩小。第二步骤是通过标准CCR模型评估子决策单元效率。第三步骤是通过子决策单元效率加权平均得出决策单元综合效率。本论文的研究主题是非同质DMUs和多流程的DEA效率评估。本论文创新性研究成果表现在以下四个方面：(1)处理非期望输出的两阶段网络DEA模型研究。近年来,雾霾气候的增多,人们越来越关注中国环境问题,不仅从经济发展的角度来看,也从人类生存的角度考虑。中国经济要达到可持续发展,管理者应该有兴趣从环境监管和经济繁荣之间找平衡点。本研究提出应用数据包络分析(DEA)测量在中国不同地区生态系统的效率,有利于管理者发现借鉴高效率的省份的成功经验。该方法不同于以往的生态系统模型,我们视生态系统作为两阶段网络流程,第一阶段是生态系统本身,第二阶段是净化系统,第一阶段的非期望输出,部分经过人工净化系统净化。净化后水回收作为一个反馈的流程。我们分开污染气体和消耗的水(被排出的污染的水)各自成为两部分,一部分是处理过的,而另一部分是未经处理直接丢弃排放到生态系统中又不能被生态系统循环和利用的污染气体和污水。第一阶段模型考虑两个期望输出,即人口和地区生产总值(Gross Region Product, GRP);以及非期望的输出变量,消耗的水和某些污染气体,例如：氮氧化物、二氧化硫和烟尘。同时,第一阶段这些非期望输出到第二阶段净化阶段作为输入。净化后的水反馈到生态系统第一阶段。因此,中间变量如消耗的水(排出的污染的水)和排放废气同时扮演既是输出又是输入双性变量角色。(2)部分输入组合在多流程中产生指定输出的DEA效率模型研究。数据包络分析是一种用来测量相关同质决策单元的效率方法。在最初模型中,假设DMUs是相同多输入、多输出指标,所有输入指标影响整体的输出指标。然而,在许多情况下这一假设并不成立。例如,在一个生产厂,包装资源(一种输入指标)只影响到需要包装的这些产品。这推断是部分的输入指标对特定输出的影响,DEA模型是基于视一个DMU内部划分为一组独立的子单元,这样DMUs的效率可以被定义为一组子单元效率加权平均。当前提出的扩展方法以允许效率测量在多个流程生成指定地输出。该模型应用于评估一组钢铁制造厂的效率。(3)非同质DEA模型关于不同的输入组合DMUs研究。在早期的研究中,学者们研究输出非同质DMUs。具体地说,不是所有的DMUs都具有相同的输出。早期的非同质DMUs,具有相同的输入指标,细分后的输出指标不同(生产产品不同)。在本章中,我们研究输入组合非同质DMUs的效率评估。以制造工厂为例,当输出组合可以使用生产机器,机器人和劳动者,因此,一个DMU输入组合不同于另一个DMU的输入组合。作为实际应用这一现象,本章分析中国各个省份的经济环境效率。所有省份都有期望输出GRP,人口和非期望输出,非期望输出例如二氧化氮,二氧化硫和空气中的粉尘等。然而,在输入端这种共性是缺失的。所有省份有水资源、资本投资和其他自然资源,后者的(其他自然资源)有几种不同的形式,即煤炭、天然气和石油。并不是所有的省份都有相同的这些资源组合,也没有明确的换算比率在这些不同资源间进行换算,但是可替换输入指标。这意味着,不能直接应用传统DEA方法。这就引出一个问题,如何公正评估输入组合不同的DMUs效率。为了解决这个问题,我们观察各个省份的不同自然资源的组合在多流程中产生输出。我们提出具有不同输入组合的非同质DEA模型,第一步骤是将DMU同质的子决策单元归为一组,对同组里的决策单元进行多流程划分,得出子单元输入和输出数据：第二步骤是利用CCR模型对同质子决策单元进行评估,得出子决策单元的效率,第三步骤通过子决策单元效率加权平均得出决策单元综合效率。通过以上三个步骤得出的DMUs综合效率,使得在评价非同质输入资源的DMU效率更为公平可信。(4)产品与副产品关联条件效率模型研究。本研究涉及部分输入输出的影响以及多流程DEA模型。早期工作没有考虑副产品在输出中的出现,由此产生的产品与副产品关联现象。模型的复杂化一方面在于产品假定两个不同的角色,作为输入指标影响副产品产生,同时作为主要输出指标。另一个复杂的方面在于多个流程出现,副产品经常出现在这些流程的一个子集中。我们开发一类基于DEA的方法模型处理在多流程中产生产品和副产品关联条件。
[Abstract]:How fair evaluation of decision making units (Decision Making Units, DMUs) relative to the peer efficiency, has been a focus of research in recent decades. Data envelopment analysis was born more than 30 years ago (Data Envelopment, Analysis, DEA) is a kind of effective evaluation of multi input and multi output relative efficiency of decision making units tools. The relative efficiency evaluation and improvement are widely used in DMUs. Many researchers have been committed to the use of DEA research, the tens of thousands of magazine articles and books published. A specific area in recent research is the utilization efficiency of network DEA measurement of DMUs. The two stage DEA is an important subset of the DEA research network, there are many studies analysis of two different stages form. One is the serial process closed, the first stage of all output as intermediate variables in the second stage, as the only input, without any elements left Study on this system. The problem is how to split the efficiency of multi stage DMUs into a single stage efficiency value. An important research achievement is combined to determine the comprehensive efficiency of DMUs by using game theory and DEA theory, and then determine the average efficiency of single stage. Then one of the important achievements in scientific research is the use of single stage efficiency geometric mean and the arithmetic average method to determine the overall efficiency of DMUs. In recent years, another important research direction is non homogeneous DMUs efficiency, non homogeneity phenomenon has a variety of forms. For example: DMUs output of different combinations and different input combinations. The first step is the comprehensive efficiency of DMUs decision-making unit through the internal structure is divided into DMUs a set of mutually exclusive, decision-making unit application and decision-making units of the same input and output indicators, the data size separated by proportion. Second steps through the standard CCR assessment model Efficiency of decision making units. Third steps through the decision-making unit efficiency derived weighted average comprehensive efficiency of decision making units. The theme of this paper is non homogeneous DMUs and multi process DEA efficiency evaluation. This thesis innovative research results in the following four aspects: (1) research on the two stage DEA network model of non expected outputs processing. In recent years, the increase in haze weather, people pay more and more attention to China environmental problems, not only from the perspective of economic development, but also from the human point of view. Chinese economy to achieve sustainable development, managers should be interested in between environmental regulation and economic prosperity to find the balance point. This study proposes the application of data analysis the envelope (DEA) measurement in the efficiency of different ecosystem China, is conducive to find out successful experience of high efficiency provinces. This method is different from the previous model of ecological system, We consider the ecological system as the two stage of the network flow, the first stage is the ecological system itself, the second stage is the purification system, non expected output of the first stage, in part through artificial purification purification system. Purified water recycling as a feedback process. We separate the polluted gas and water consumption (by the discharge polluted water two) each become part of a part is processed, and the other part is untreated discarded directly discharged into the ecosystem can not be polluted gas and sewage ecosystem circulation and utilization. The first stage model considering two output, population and GDP (Gross Region, Product, GRP); and non expected output variables, the consumption of water and some polluting gases, such as nitrogen oxides, sulfur dioxide and smoke. At the same time, the first stage of these undesirable outputs to the second stage purification stage As the input. The purified water back to the first stage of ecological system. Therefore, the intermediate variables such as consumption of water (water pollution) emissions and output is also play both input variables. The role of bisexual (2) of DEA efficiency model are combined to produce a specified portion of the input output in the process of data envelopment. The analysis is a measure of the efficiency of decision making units related to homogeneous method. In the first model, assuming that DMUs is the same as multi input, multi output index, the output index overall all input index. However, this assumption does not hold in many situations. For example, in a factory for the production of packaging resources (a input indicators) only affect the need for packaging of these products. This inference is the input index influence on the part of the specific output, the DEA model is a DMU based on the partition for a group of independent sub units, so the efficiency of DMUs Can be defined as a group of sub unit efficiency. The weighted average method is proposed to allow the extension of the current generation of specified output efficiency measurement in multiple processes. This model was applied to evaluate the efficiency of a group of iron and steel factory. (3) non homogeneous DEA model on input combinations of DMUs. In the early studies, the scholars study output non homogeneous DMUs. specifically, not all DMUs have the same output. Early non homogeneous DMUs, with the same input indicators, output indicators after the breakdown of the different (the production of different products). In this chapter, we study the input combination evaluation efficiency of non homogeneous DMUs. Manufacturing plant for example, when the output you can use a combination of production machines, robots and workers, therefore, a combination of input DMU input combination is different from another DMU. As an application of this phenomenon, this chapter analyzes the various provinces by Chinese Economic and environmental efficiency. All provinces have the expected output GRP, population and non expected output, non expected outputs such as nitrogen dioxide, sulfur dioxide and dust in the air. However, at the input end of the common deletion. All provinces have water resources, capital investment and other natural resources, the latter (natural resources) there are several different forms, namely coal, gas and oil. Not all provinces have the same combination of resources, there is no explicit conversion ratio conversion in these different resources, but can replace the input index. This means that the traditional DEA method can not be applied directly. This leads to a problem how, impartial assessment of the different input combinations of DMUs efficiency. In order to solve this problem, a combination of different natural resources we observe various provinces to generate output in many processes. We present a different input group Non homogeneous DEA model, the first step is the decision-making unit of DMU homogeneous are classified as a group, multi process classification for decision making units in the same group, the sub unit of input and output data: the second step is to evaluate the same proton decision unit using the CCR model, the efficiency of decision-making unit, third steps through the decision-making unit efficiency of decision making units that the weighted average comprehensive efficiency. Through the comprehensive efficiency of DMUs above three steps that make it more credible in the evaluation of non homogeneous input resource efficiency. DMU (4) contact conditions study efficiency model of products and by-products. This research involves part of input and output and the influence the process model of DEA. Early work did not consider the products appear in the output, resulting in the phenomenon of associated products and by-products. A complex model is a product of two different assumptions The role, produced as input indicators affect products, also as the main output index. Another aspect is a complex process, products often appear in a subset of these processes. We developed a method based on the DEA model in multi process produces products and by-products related conditions.

【学位授予单位】：中国科学技术大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：F224

【相似文献】