It's been a couple of years since I've done anything with neural networks, but I think I shifted away from using NARX altogether. So, unfortunately, this is still a mystery to me. There's an answer below from Greg, if you'd like to try that. Good luck!
Multiple input NARX?
5 views (last 30 days)
Show older comments
I've been thinking about purchasing neural network toolbox, because I think NARX could be the right approach to my forecasting problem. However, it seems that using multiple inputs has been a big issue, since googling "NARX" gives numerous topics where solutions have been requested but never answered. However, most of these topics seem to be several years old.
My question is this: is it possible to include multiple (say, three) exogenous inputs to train NARX? I found out that MATLAB has a function called tonndata. Is this a new function designed to do that? Could I take my three inputs of form
x1(t1) x2(t1) x3(t1)
x1(t2) x2(t2) x3(t2)
x1(t3) x2(t3) x3(t3)
where t:s are there to indicate time points and x1, x2 and x3 are some sort of variables (say, current, temperature and velocity), feed this matrix into tonndata and then be able to feed the outcome to NARX? Or is there some other way of providing multiple input variables?
Answers (1)
Greg Heath
on 17 Dec 2016
Edited: Greg Heath
on 17 Dec 2016
Vector inputs and targets are relatively easy to use because the MATLAB equations are in matrix form.
If you search for nonscalar example datasets use
help nndatasets
and
doc nndatasets
I have enumerated the dimensions of the inputs and targets in three NEWSGROUP posts:
http://www.mathworks.com/matlabcentral/newsreader/view_thread/...
337914
338208
339984
Thank you for formally accepting my answer
Greg
1 Comment
See Also
Categories
Find more on Sequence and Numeric Feature Data Workflows in Help Center and File Exchange
Products
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!