Dear Parsapoor

Thanks for answering my question

I have gone through the link provided by you. In the example provided there, the pH of a solution is predicted provided the flow rate and other input time-series data are available. My problem is that I have only one time series available and I want to forecast the future values of this time series . Though other neural networks such as NAR and time-delay networks in MATLAB toolbox are able to forecast ,if only one series is available, I want to know that is it possible for a NARX network also. 

ANFIS and some other time-series prediction techniques such as support vector regression are able to forecast but I am pretty much interested in the NARX model . But the BELFIS technique sounds interesting and is new to me. Could you please tell me the software in which I can implement BELFIS. Is it possible to execute it in MATlAB.

In the matlab neural network toolbox I think you're looking for narnet (link attached).  This is basically the same as NARX but without the 'X' which represents the external input time series.  This model will predict future values of a time series using only a time window of values of the same time series.

http://www.mathworks.com/help/nnet/ref/narnet.html

Dear Matthew,

I agree with your view but I have found certain research papers where NARX network is being used for time series prediction. As an example please go through the paper.

Menezes, J. M. P., & Barreto, G. A. (2008). Long-term time series prediction with the NARX network: an empirical evaluation. Neurocomputing, 71(16), 3335-33

I also feel what u r saying is true. Please inform me if you get any idea after reading this paper

To train the neural network with this capacities, during the training process, tghe data must be set in order. So, the final system consider the time as a part of the processing.

It is important to use also as activation function sigmoids