数据包络分析是一种评价具有多投入、多产出决策单元的相对效率的线性规划方法.在现实世界中,决策单元有时呈现出由多个独立子系统构成的复杂并联网络系统,各子系统的投入/产出之和构成了系统的总投入/产出. 目前,用于评价这种具有并联网络生产系统相对效率的模型主要有三种:网络 DEA 模型、多部门 DEA 模型和关联 DEA 模型.现有这些模型的基本特性和相互关系存在着不足,即子系统的效率分解和优化指数不唯一. 为解决这一问题,提出了改进的并联 DEA 模型,并采用加拿大银行系统实例来说明所提出模型的合理性和有效性.

Data envelopment analysis (DEA) is a linear programming approach for evaluating relative efficiency of peer decision making units (DMUs) that consume multiple inputs to produce multiple outputs. DMUs can have parallel independent sub-units, where inputs and outputs of each DMU are the sum of those its sub-units. This paper investigates the properties and relationships of the existing parallel DEA models, i.e., network, multi-component and relational DEA models, that address measuring the performance of parallel production systems. A major limitation of the existing DEA models is that efficiency optimization may produce multiple sets of efficiency scores for individual sub-units. The current paper proposes a new DEA approache through efficiency decomposition to deal with this problem. A case of Canadian bank branches is employed to illustrate these parallel DEA approach.