Traditional DEA method considered decision making units (DMUs) as a black box, regardless of their internal structure and appraisal performance with respect to the final inputs and outputs of the units. However, in many real systems we have internal structure. For this reason, network DEA models have been developed. Parallel network DEA models are a special variation which inputs of unit allocated to several sub-processes, and the total output of these processes are the output of the relevant unit. Parallel network DEA model have been developed based on certain and confidence data and they do not have the ability to deal with uncertainty about inputs and outputs data. In this paper we are going to provide a robust parallel network DEA model and attempt to describe and analyze the new model with a real case study.
The proposed model can evaluate the performance of parallel structure and it has the ability to deal with uncertainty in the data. The results show that with increasing levels of uncertainty and decreasing of reliability, efficiency of decision-making units will be further reduced. Increasing of discrimination power is another advantage of the proposed model. Also traditional network models show more deviations in the results in compare with robust network model when the data are changed.

Keywords: Efficiency, Network DEA Models, Robust Optimization

Full-Text [PDF 780 kb]   (923 Downloads)