Blog post on using LSTM for prediction in R

[Rud854]

Seeing your book, congratulations on the excellent work you have done I didn't see the use of LSTMs in tidyfinance. LSTMs are pretty useful tool for forecasting in finance. I suggest using R torch or mlr3torch. A dedicated chapter could work too. i use LSTMs for some finance prediction in Python (not R since there isn't much resources ) though I won't specify.

[christophscheuch]

Thank you for the suggestion. Could you point us to specific papers or blog articles that you find useful in this context? Frankly, I have not yet worked with LSTMs at all.

[Rud854]

https://www.sciencedirect.com/science/article/pii/S0950705125006094
It was based on this paper, it seemed interesting to even combine ARIMA with LSTMS. The method I currently know is to use ARIMA and then for residuals to use a LSTM. But you could expand the concept with other models. The paper I think was done in Python but it would be cool to expand in R too.

[christophscheuch]

Thank you for sharing the reference. The topic looks like a good candidate for a blog post. Unfortunately, we currently have limited capacity to prioritize this on our end in the near future. Would you be open to contributing a simple application of LSTMs using Python and PyTorch? I’d be happy to provide guidance and review your work.

[Rud854]

I think I'm willing to, because I got a lot of data science tasks ahead of me for work it'll take a few weeks or 1-2 months to focus on this. I might do a ARIMA- LSTM hybrid in R I know it's possible if I use forecast and R torch as well.