考虑系统要素相关性的DEA效率评价方法及在商业银行中的应用研究

As evidenced by the practice, due to the contagion effect of the correlation network among organizations, the imperceptible changes of the system element often tend to have great influence on the performance of organizations. As a result, it is urgent to consider the correlation effect of system elements in efficiency evaluation. As an important tool for measuring the relative efficiency of decision making units (DMUs), data envelopment analysis (DEA) has made great progresses both in methodology and application research. Weights, indicators, and DMUs are the three basic elements of DEA model, and are also typical factors of influencing the performance evaluation for organizations. According to the three elements, the project focuses on three typical correlation scenarios to make DEA efficiency evaluation, namely weights correlation, indicators attribute correlation, and DMUs correlation. Then, based on the possibility of realization of the efficient front, the instability of the reference set, and the dynamic relationship among DMUs that result from the correlation scenarios, the classic DEA theory on efficiency evaluation is systematically extended to network correlation efficiency evaluation. The efficiency evaluation method based on the three correlation scenarios is built by means of using DEA, mathematical statistics, game theory, algorithm design, and data mining, etc. Finally, the proposed method is applied to commercial banks to make decision on pricing of financial assets, performance improvement, and competition strategy. These research will further promote the DEA research from theoretical area to broader application area, and provide valuable decision information and theoretical suggestions to decision makers.

实践多次证明，通过相关性网络的传染效应，系统要素的细微变化常会给组织效率带来巨大的影响，因而在组织效率评价中迫切需要考虑到系统要素相关性的影响。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。权重、指标、DMU构成了DEA模型的基本要素，也是组织绩效考核的典型影响因素。因此本项目将系统研究权重、指标属性和DMU关系三种典型相关性情形下的DEA效率评价方法。其次，基于各种情形引起的有效前沿面可实现性、参考集不稳定性及DMU间动态关联性等，将DEA理论系统地向网络相关评价拓展；再利用DEA、数理统计、博弈论、算法设计、数据挖掘等方法，构建三种相关性情形下的DMU效率评价方法。最后，以我国商业银行为背景开展金融产品定价、绩效改进、竞争策略等应用研究。研究将进一步推动DEA研究从理论层面走向更广阔的应用层面，为决策者提供有价值的决策信息和理论借鉴。