In order to respond effectively to his clients, the National Gas Company needs to know about the amount of prevailing demand in every period. From the other hand, lack of information about the amount of subscription demand will in turn cause certain difficulties such as not knowing the number of required contractors, the absence of an appropriate inventory control for all kinds of needed gas consumption counters, etc. In the last few decades, economists and management specialists have frequently used econometric methods for estimating demand. Today, among the available estimation methods, they are using the artificial neural networks methods and fuzzy models in many areas of applications and each of them have its own merits and limitations. A successful combination of these two methods, with emphasis on learning power of neural networks and logical performance of fuzzy model, have been transformed into a very powerful instrument that are now being applied in different contexts. The main purpose of this article is to estimate the urban household gas subscription demand by using the ARIMA linear method and the fuzzy neural networks nonlinear method and to compare them on the basis of six performance criteria. The article concludes that, considering the six performance criteria, the fuzzy neural networks nonlinear method has supremacy over the ARIMA linear method for estimating the urban household gas subscription demand; therefore, it is more appropriate.