Is there an R tutorial of using LSTM for multivariate time series forecasting?

Question

There is a great blog post about how to use keras stateful LSTM in R to forecast sunspots. I applied it to financial ts data sets, and the problem I am encountering is that some of the forecasting numbers are way off to a degree that are unreasonable.

Therefore, I am wondering if there is an R tutorial of using LSTM for multivariate times series forecasting? I'd like to include variables like opening and closing price because I think that will "normalize" the forecasting values. I found a few tutorials in Python but I have limited experience with it. Maybe it's a time to pick up Python? Any advice or suggestions will be appreciated!

Answer 1

I like the Jena data example. In multivariate settings, you only need to generate lookbacks over all X.

https://blogs.rstudio.com/tensorflow/posts/2017-12-20-time-series-forecasting-with-recurrent-neural-networks/

Answer 2

I'm dealing with the same issue. I know this post is pretty old but I have found some recent posts that may help: there is a 3-part series on this subject in R studio AI blogs.

https://blogs.rstudio.com/ai/posts/2021-03-10-forecasting-time-series-with-torch_1/

It's not perfect, but it answered enough for me to make forward progress on the multivariate complexities.