Using of NARX in the case a time series step is not constant and inconsistent from one sample to another.

I have 200 samples,each sample has 4 Inputs constant except time(a,b,c,(t1,t2,t3))and 1 Output (variable according to the t1,t2 and t3).The(t1,t2 and t3)is not consistent for all samples i.e. for first sample(t1=1 day,t2=40 day,t3=180 day),Second sample(t1=5 day,t2=55 day,t3=195 day) and so on..... My question: Is Narx(dynamic neural network) benefit to solve my problem and if not what is a appropriate architecture of ANN that i should use.I note that i use Toolbox of Matlab not a programming code.

Answer 1

[Greg Heath]

Terminology: You have 1 sample of 200 examples.

Do you mean you have 3 constanr inputs (a,b,c) ?

If so, ignore them and you will have

 [ 3 200] = size(input)

 [ 1 200] = size(output)

Unless there is a relatively constant correlation between the current output and past inputs and/or outputs, you have a simple regression problem. See

 help fitnet
 doc fitnet

Hope this helps.

Thank you for formally accepting my answer

Greg

[Abdulaziz]

Thank you for reply .i load my problem as a image, as you see i have (200X3=600 output and (3(inputs)*200+600 input as time).The time step id different for each sample and the Output is increase whith time.What are you advice me?

Answer 2

[Greg Heath]

 [ 4 600 ] = size(input)
 [ 1 600 ] = size(target)

 help fitnet
 doc fitnet

Answer 3

[Abdulaziz]

Thank you very much Greg Heath for all your help.