Efficiency Estimation using Nonlinear Influences of Time Lags in DEA Using Artificial Neural Networks

One of the common methods for the assessment of an organization's efficiency is comparison with other competitors. However, some researchers have studied the efficiency of a unit within itself during different periods of time and it is used to investigate the performance trend of the unit during previous times. The purpose of this research is to forecast the performance of a unit using the previous time series of its performance. This research conducts comparison and efficiency analysis of a unit during different time periods using SBM and DEA models. And then, the outcome is considered as the training elements of an ANN, so efficiency of future time steps can be estimated for that unit. An industrial unit in Steel industry was studied in this research and its decreasing performance trend during ten years has been presented after efficiency measurements. Implementing different structures of ANNs, finally, we found out that a recurrent neural network with 10 neurons in a hidden layer and Bayesian Regularization algorithm had the best performance for future forecasting of efficiency. The most important achievement of this study is efficiency forecasting for organizations' future using the existing data with regards to the influences of previous time steps on current efficiency by a nonlinear approach. It would lead to providing a clear image of the organization's future as represented for the case of this paper.