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

See more:

network

neural-network

I have 200 samples,(In the bottom link is a obvious sketch)each sample has 4 Inputs (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=3 day,t2=7 day,t3=15 day),Second sample(t1=2 day,t2=9 day,t3=17 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 want to prediction of output with inputs within this range not in future.I note that i use Toolbox of Matlab not a programming code.

http://www.mathworks.com/matlabcentral/answers/uploaded_files/14084/test.png