Technical Efficiency Analysis of Oil Refineries Using Combinational Model of Neural Networks and Data Envelopment Analysis (Neuro-DEA)

In today's volatile environment, performance evaluation is a critical problem. In recent decades, various models evaluating performance have been proposed but the model standing out is Data Envelopment Analysis (DEA) with its own advantages and disadvantages. One of the most considerable problems that limit DEA performance is the small number of Decision Making Units (DMUs) in proportion the number of inputs and outputs. When the aforementioned problem exists, basic DEA models can not efficiently differentiate units and have difficulties tackling it Hence, the differentiability power of the model is low. In this paper Artificial Neural Networks (ANNs) are used to measure technical efficiency and a combinational model of ANNs and DEA (Neuro-DEA) is proposed. Also the topology of the proposed model which is a two layer perception with back propagation algorithm is analyzed. Furthermore in a case study related to the evaluation of oil refineries technical efficiency, results of Neuro-DEA are discussed and compared with input oriented CCR model.