I have an IT background and I would like to use data from a forex calendar like this one to predict prices.

The problem is that calendar news impacts can last for days or weeks or even can effect previous hours to their release.

Since I would like to use this fundamental information as signal for a trading system, I have read about Time Delay Neural Networks and I think it could be useful for this task.

Does anyone knows or can point to references on how can I apply a Time Delay Neural Network to accomplish this? Thanks beforehand!


I suggest you to read as reference the following paper:

Yao, Jingtao, and Chew Lim Tan. "A case study on using neural networks to perform technical forecasting of forex." Neurocomputing 34.1 (2000): 79-98.

The authors applied TDNNs to the forex markets and I think it could be useful for you purposes; you can find the pdf version of the paper here.

As regards the available tools to construct this kind of model, I know you can implement that in matlab, by exploiting `timedelaynet.

Hope this will help.

  • $\begingroup$ Many thanks! I have been searching for a R TDNN package with no sucess, do you know any? $\endgroup$ – MithPaul Jul 5 '15 at 23:44
  • $\begingroup$ I do not know if there exist, but, honestly, I think not; I follow r-bloggers.com and I'm pretty informed about NN field and I know there are mainly two packages about NN: 'nnet' and 'neuralnet'. You should try to adapt those packages to your needing! If you're satisfied about the answer, please mark that, thanks @MithPaul! $\endgroup$ – Quantopik Jul 5 '15 at 23:49
  • $\begingroup$ Thanks again @Quantopic, the closest I found is RSNNS package for R: cran.r-project.org/web/packages/RSNNS which has implementations of Elman and Jordan Neural Networks which are both Recurrent Neural Networks. I think that for my purpose they are the same as a Time Delayed Neural Network (but I realise the question about the differences between them would suit best to stats.stackexchange) $\endgroup$ – MithPaul Jul 6 '15 at 0:57

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.