企业综合经营效率的两阶段评价研究

【作者】 李涛；

【导师】 高阳；

【作者基本信息】 中南大学 ， 管理科学与工程， 2008， 硕士

【摘要】 效率是经济学研究的核心问题之一,也是公司竞争力的集中体现。因此,要全面提升公司的竞争力,应对中国市场全面开放和外资公司涌入给我国企业带来的巨大压力和挑战,就必须高度重视效率问题。高效率作为企业努力追求的目标之一,日益成为关注的焦点。对企业的综合经营效率分析正是围绕企业的生存和发展能力给予客观评价。本文秉承这样一个思路:将对企业综合经营效率的评价分成两个阶段:第一阶段从投入产出的角度来来判断企业经营效率的相对有效性,对于经营效率相对有效和弱有效的单位,再进行第二阶段的评价,即对这些综合经营效率相对有效和弱有效的单位进行评价排序,最后得出每个企业在所有被评价单位中的经营效率排名。依据这样的思路,本文建立了两个模型,分别用于两个评价阶段。并通过实例分析,对两种方法的有效性进行了分析。具体而言,主要工作及成果如下:(1)建立了一种基于偏好约束锥的SO-DEA模型。在建模过程中,将熵权法和离差最大化法运用到指标客观权重的求解中;根据矩估计的思想,通过建立优化模型並经专家打分得到的主观权重与客观权重集结成综合指标权重。利用得到的综合指标权重借鉴AHP方法,建立指标判断矩阵,同时根据该指标判断矩阵建立了偏好约束锥,最后在此基础上建立了基于偏好约束锥的SO-DEA模型。该模型主要用于企业综合经营效率评价的第一阶段。(2)建立了一种基于量子粒子群算法的神经网络模型。该模型将量子粒子群算法和神经网络相结合,使其既具有神经网络的广泛映射能力,又具有量子粒子群算法带来的高效率,全局收敛性等特点。该模型主要用于企业综合经营效率评价的第二阶段。(3)通过实例分析论证了所提出方法的有效性。针对第一阶段的评价,分别用传统的C~2GS~2模型和本文建立的SO-DEA模型对国内某行业的其中67家上市公司进行了总体评价,结果显示利用SO-DEA模型评价确实体现出了偏好约束锥的约束作用。该阶段得到了综合经营效率相对有效和弱有效的25家单位。针对第二阶段的评价,本文用经典粒子群算法和量子粒子群算法分别训练神经网络,发现基于量子粒子群算法的神经网络在保证训练速度的前提下,其全局搜索能力更好,而且训练误差更小。通过matlab7.0仿真计算,对综合经营效率相对有效和弱有效的单位进行了具体排序。更多还原

【Abstract】 Efficiency is one of the most important area in economics, it is also a concentrated expression of corporation’s competitive strength. Therefore, in order to completely improving coporation’s competition,to cope with the severe challenge and terrible pressure from the full liberalization of China’s markets and the inburst of foreign company, efficiency problem must be attached to great importance. As one of the goal pursued by corporations, high efficiency becomes the center of attention increasingly.To analysis corporations’ integrated efficiency is to give an objective evaluation on the ability of existence and development.The paper is based on the follow train of thought:in order to analysis corporations’ integrated efficiency in one industry, decision makers can firstly validity on the whole,then they can arrange all the corporations by integrated efficiency values.In this way,we can tell the level of every corporation’ integrated efficiency. according to this train of thought,the paper builded two mathematical model to apply to the two-phase process. in addition, the applicability of the two methods are analysised through case-study.It should be noted that the paper concentrates on the study of evaluation method, doing little work to address how to choose index.In this regard, the main achievement are as follows:(1) A kind of DEA model based on cone ratio is created. According to moment estimate requirements, index weight vectors are obtained by combining the resuts of maximizing deviations method and entropy weight method.Making use of the integrated index weigh,the paper creates index judgment matrix which is used to build the cone ratios refer to AHP method.(2) Creating an artificial neural network based on Quantum-behaved Particle Swarm Optimization,the paper combines QPSO with ANN,making the network have not only extensive mapping capability but also high efficiency and global convergence.It also analysises its applicability through case-study.(3) The paper proves the applicability of the two methods.For the first method,the paper evaluates the integrated efficiency of 35 quoted companies in a specific industry with classical C~2GS~2 model and SO-DEA modelseperately. For the second method,the paper makes use of classical PSO and QPSO seperately to train neural network,finding that the later method is better than the former in training speed, global search,and precision.更多还原