动态环境下决策单元效率评价方法与应用研究

A real-time and continual performance evaluation for organizations is the precondition to response the turbulent decision-making environment quickly and to achieve a dynamic management effectively. As an important relative efficiency assessment tool for decision making units (DMUs), to this day, data envelopment analysis (DEA) has gained a great development on both the theory and application research. However, with the turbulent decision making environment, how to properly separate decision environmental effect from DMU operations and how to respond to the DMUs' numbers change and their relationship reorganization, has become the new challenges for traditional DEA methodology. Therefore, the project firstly focuses on three typical dynamic decision environment, namely dynamic operations environment, dynamic data environment and dynamic relationship environment, and decomposes and integrates each possible circumstance in detail. Accordingly, with considering DMUs' non-homogeneity, the instability of reference sets and DMUs' relationship reorganization, the classic DEA theory is extended from static efficiency assessment to dynamic efficiency assessment. Based on the dynamic DEA theory, three types of methods of DMU's efficiency assessment with DEA method, mathematical statistics, algorithm design and data mining, are proposed under the above mentioned dynamic decision-making environments. Finally, these models and methods will be applied to the efficiency assessment, the industrial policy effects estimates and the industrial space optimization of retail industry in China, which will be helpful for improving the timeliness, sustainability and feasibility of Chinese future policy making in retail industrial. The value of the project lies in that it can enrich DEA theoretical system, extend DEA application, and offer decision-making support for industrial policy planning, simultaneously.

实时持续的绩效评价是响应动态决策环境、实施动态管理的首要前提。作为评价决策单元（DMU）相对效率的重要工具，数据包络分析（DEA）方法至今已获得极大的发展，但面对动态的决策环境，传统DEA方法仍然面临合理分离环境效应、快速响应DMU数量增减以及DMU间关系重构等诸多挑战。为此，本项目相应考虑三种典型动态决策环境：动态运营环境、动态数据环境与动态关联环境，并对三种环境的可能波动情境进行细分；其次，基于各种动态情境引起的DMU异质性、参考集的不稳定性以及DMU间的动态关联性等，将传统DEA效率评价理论由静态评价向动态评价拓展；再运用DEA、数理统计、算法设计、数据挖掘等方法，构建三类典型动态环境下的DMU效率评价方法。最后，运用上述方法对我国零售业发展绩效进行实时评价，据此提升未来产业决策的实时性、持续性与可行性。可见，本项目研究将进一步推动DEA 方法从理论层次走向更为广阔的应用层面。

实时持续的绩效评价是响应动态决策环境、实施动态管理的首要前提。作为评价决策单元（DMU）相对效率的重要工具，数据包络分析（DEA）方法至今已获得极大的发展，但面对动态决策环境与复杂生产运营系统，传统 DEA 方法仍然面临多阶段生产运营冲突与协调、动态多期资源配置以及环境动态实时评估等诸多挑战。为此，本项目重点研究了如下问题：1）静态环境下串联与并联生产系统的效率评价与前沿投影研究；2）动态环境下生产平稳性与技术异质性资源配置研究；3）动态环境下环境效率评价研究；4）DEA效率连续性测度与离散数据DEA效率分析。主要研究结果和贡献包括：1）提出了可变松弛变量测度模型，解决了传统模型前沿面投影及和无法分解阶段效率的问题；基于钢业产业循环用水评价，建立了考虑非期望产出的两阶段DEA模型，丰富了两阶段DEA方法体系；建立了并联生产系统（多部结构）的评价方法，提出了刻画方法差异的两种一致性条件，解决了传统方法可能会受到多重最优解困扰的不足；2）建立了动态环境下考虑生产平稳性的排污权配置DEA模型，丰富了现有污染气体排放权分配方案，完善了排放权分摊的应用实践；提出了技术异质性动态环境下的集中化资源分配和成本分摊模型，进一步拓展了现有资源配置DEA方法研究；3）构建了meta前沿面和群组前沿面DEA模型，刻画了城市群的生产技术差异，提高了城市群环境绩效评价的准确性；4）针对DMU快速增加或合并情形，提出了基于松弛变量测度的联合DEA模型，保证了效率评价的连续性和Pareto有效性。总之，上述研究不仅丰富了现有DEA理论体系，相关应用研究也能为复杂生产系统绩效评价与改进、动态环境绩效评价与资源配置决策提供有价值的决策支撑。