Super efficiency SBM-DEA method and neural network for the efficiency evaluation in the case of data uncertainty
Subject Areas : International Journal of Data Envelopment Analysis
1 - Department of Mathematics, Science and Research Branch, Islamic Azad University, Tehran, Iran
Keywords: Neural Network, Efficiency Evaluation, Super-efficiency SBM, Data Envelopment Analysis,
Abstract :
The classic models for the performance assessment in Data Envelopment Analysis (DEA) may have some inherent issues. For example, they can be affected by the statistical noise in data. Furthermore, if the decision maker (DM) adds new decision-making units (DMUs) into the evaluation, then the performance of all the original units is affected and must be re-measured, which restricts the efficiency evaluation in DEA. The main goal of this paper is to apply machine learning algorithms to overcome the shortcomings of the DEA models. On the other hand, in many real-world problems, there are some imprecise data due to incomplete or non-attainable information, errors in measurements, unquantifiable variables, or any other source of reason. In the DEA literature, many studies have focused on developing methods that incorporate uncertainty into the input/output values. The uncertain data can be reported as fuzzy data, stochastic data, and interval data. This paper considers the situation where each input/output value is selected from a symmetric box. First, we use a super-efficiency SBM model in the presence of uncertain data to construct the relative effective frontier and then apply the machine learning algorithms to construct a regression model and establish the absolute effective frontier. The proposed method has some advantages, compared to the existing methods. Also, the proposed model can better overcome the problems associated with DEA compared with the DEA in the presence of uncertain data and the neural network fusion outlined in the literature, so it can improve fusion efficiency.