Model improvement for computational difficulties of DEA technique in the presence of special DMUs

Data envelopment analysis (DEA) is a non-parametric mathematical optimization technique which evaluates the relative efficiency of decision making units (DMUs), with multiple inputs and outputs. A major advantage cited in support of the use of DEA in measuring efficiency is that this method does not make assumption about functional forms of production function. In other words, it makes a piecewise frontier (efficient frontier) with calculation of a maximal efficiency measure for each DMU relative to all other observed measures. Also, it identifies a subset of efficient “best-practice” DMUs and for the remaining DMUs, the magnitude of their non-efficiency is measured by comparing to a frontier constructed from the efficient DMUs. Therefore, in the evaluation of non-efficient units by DEA, referenced DMUs have an important role to play. Unfortunately, DMUs with extraordinary output can lead to a monopoly in reference set. The fact that may call it abnormality due to outliers data. In this paper, we introduce a DEA model for evaluating DMUs under this circumstance. Also, the model is employed to measure the bank branches efficiency of a large Canadian Bank. The layer model can derive a ranking for DMUs and obtain an improving strategy for moving toward a better layer.