数据包络分析（DEA）是对决策单元进行效率评价和改进的重要研究方法，而含有非期望产出的决策单元效率评价更是DEA的热点研究领域。该领域现有的DEA理论方法对不同非期望产出性质刻画较为简单，且忽略了它们之间的关系，无法有效解决现实中具有非期望产出复杂特性的决策单元效率评价问题。本项目以非期望产出复杂特性为切入点开展如下研究：1）从现实问题中抽象出几种关键的非期望产出复杂特性表现，将其刻画分类为多样性、关联性及异质性特征，研究其对效率评价和效率改进的影响机理；2）分别研究考虑每一种特征的DEA效率评价与效率改进方法，构建系统的DEA模型和方法体系，并论证其可行性和科学性；3）将模型和方法应用于运输业效率评价及非期望产出治理，以验证其有效性。本项目将对现有DEA体系进行补充和完善，推进DEA理论方法发展和应用实践，并为现实中考虑非期望产出复杂特性的决策单元效率评价、效率改进和管理决策提供支持。

Data envelopment analysis (DEA) is a very important tool for efficiency evaluation and improvement of decision-making units (DMUs). Efficiency assessment considering undesired output is a research focus of DEA. The existing literature of DEA methods with undesirable outputs involves few types of undesirable outputs and rarely considers the relations among undesirable outputs. It results in that the efficiency evaluation of DMUs with complex undesirable outputs in reality cannot be solved. In this project, the complexity of undesirable outputs is considered. Firstly, we define three main characteristics of undesirable outputs complexity as diversity, interdependency and heterogeneity, and study the impact of these characteristics on the efficiency evaluation and efficiency improvement. Secondly, a set of DEA models will be constructed for evaluating efficiency of DMUs with each characteristic, of which the feasibility and effectiveness will be demonstrated. Finally, the proposed models and methods will be employed to an empirical study on transportation efficiency evaluation with complex undesirable outputs to verify their effectiveness. This project will improve the current DEA methods to be applied in real situation, and provide supports for assessing the efficiency of DMUs considering complex undesirable outputs.