Proposing a Mixed Model Based on Stochastic Data Envelopment Analysis and Principal Component Analysis to Predict Efficiency

Data Envelopment Analysis is a technique based on linear programming methods to construct surface or non parametric boundary over the data. This boundary is used to evaluate proportional efficiency. In this paper, a mixed model was proposed based on Stochastic Data Envelopment Analysis (SDEA) and Principal Component Analysis (PCA) for predicting the similar units' efficiencies in an organization. It was tried to abate the most important shortcoming of DEA, involving: being unable to estimate the efficiency, the unreal distribution of weights of inputs and outputs of the model and variety in efficient branches. The following model covered the mentioned problems caused by entering the stochastic effect, and considering the effects of fuzzy weights on the inputs and outputs by the use of SDEA technique with fuzzy weights. PCA was used to determine efficiency mean for units with various risks. Finally, in order to reach a better understanding of the proposed model, it was applied to predict efficiencies for some Iranian Bank branches. The high correlation between real and predicted efficiencies was obtained which represented the validity of the proposed SDEA/PCA model.