I've heard a bit about using neural networks to forecast time series, specifically recurrent neural networks.
I was wondering, is there a recurrent neural network package for R? I can't seem to find one on CRAN. The closest I've come is the nnetTs function in the tsDyn package, but that just calls the nnet function from the nnet package. There's nothing special or "reccurant" about it.
There is the RSNNS package that provides access to the "Stuttgart Neural Network Simulator" (SNNS). It contains the classical recurrent network structures of types 'Jordan' and 'Elman'. SNNS is a bit old (before 2000), but may still be worth a try. The R package itself has been updated in September this year.
I am hoping someone with more R knowledge than me will submit an R answer, but I am not aware of anything. Here is one option: Use one of the multiple Python-based implementations (e.g. PyBrain or PyNeurGen) and interface back to R via Rpy or (my prefence) pyRserve. I know this is not ideal, but it could given you an easier way forward than writing your own package, at least at first. Also, I am guessing it would be preferable to call Python from R, but I don't think the RSPython package in R has been updated for some time.
EDIT: It looks like PyNeurGen may not have been updated in some time either. PyBrain seems to have the largest following and is under active development.
There is a new package out: rnn (on CRAN, on github), which implements a recurrent neural network in native R code.
A nice example can be found here:
http://firsttimeprogrammer.blogspot.de/2016/08/plain-vanilla-recurrent-neural-networks.html
