Using Artificial Neural Networks in Efficiency Measurement of Banking Industry (Case Study: Tejarat Bank Branches in Mazandaran Province)

Modern techniques for measuring performance have divided into two areas of parametric and non-parametric borders and each of them has special capabilities and limitations. Meanwhile, DEA, i.e. data envelopment analysis, is the most common and most widely used approach. However, this technique has some shortcomings such as complexity in determining the type of returns to scale and the deviation of the border if there are random effects. To deal with these problems, the approach of using neural networks to estimate the efficiency frontier is helpful. Because this view requires no special default, and estimates the production functions with high precision and has the ability to operate in non-linear contexts. Therefore, in this study, the neural network capability in estimating the efficiency frontier compared to the DEA efficiency frontier. For this purpose the performance of 85 branches of Tejarat Banks, in Mazandaran province, were calculated and compared using neural network and data envelopment analysis (CCR and BCC models). Based on neural networks, just the performance of the central Amol branch was 100 percent and DEA also recognized this unit as efficient. Comparisons revealed that a similar way exists in both strong and weak units and there is a strong correlation relation between rating and performance scores provided by the two methods. At the end, Improvements in the efficiency of inefficient branches based on the border of the neural network were introduced.