One of the biggest problems of using Data Envelopment Analysis for the evaluation of performance is the weakness of the separability for the decision maker units. This problem is generalizable because of the lower quantity of units in comparison with the ¬input and output quantities of the model . This matter considerably reveals itself in the evaluation process of 23 provincial gas companies considering the higher input and output rates of each provincial gas company. Based on this fact, in this research an integrative model composed of Performance Predictor Neural Networks and Data Envelopment Analysis is designed and used. The results using this model illustrate the strength of the neural network in evaluation and classification of the companies based on their efficiencies.