Performance Evaluation of Decision Making Units Using Data Envelopment Model and Artificial Neural Network (Case Study: Fars Regional Water Corporation)

Abstract

One of the problems with using the DEA is the lack of resolution for decision makers. The performance limit obtained by the DEA is also sensitive to statistical perturbations and outliers caused by measurement error or any other external factor, causing the efficiency limit to be shifted and diverting the DEA analysis path. The DEA can also hardly predict the performance of decision-making units in the future. Therefore, artificial neural networks are a good tool to use in such issues because the nature of ANN performance is due to its learning power and generalizability in a way that is more resistant to outliers and perturbations resulting from inaccurate data measurement and can As a useful tool for managers to predict and observe the behavior of their system in the organization in the future. Also, in order to implement the theoretical findings from practical research, 27 district units of Fars Regional Water Company to increase the volume Groundwater was evaluated. Initially, the input-driven CCR model and the Anderson-Peterson (AP) method were used to rank the units in the DEA model, and then the ANN approach was used to evaluate the performance of the units using the hybrid models (DEA - Neuro). The results of the computational efficiency analysis of the units using this model demonstrate the high power of the network in computing and resolving the performance.