search

cchallu / nbeatsx

MIT License
185 stars 48 forks source link

Forecasting with long output dimension #6

Closed linotto closed 2 years ago

linotto commented 2 years ago

Hello,

First of all thank you for your paper, I have tried N-beatx and it's performing super well.

The problem is that I need to forecast not only Day-Ahead but seven days (so 168 hours) . I tried to put my data in daily and did a 7 forecast horizon but its not good, do you have any advice to try for that problem please ? I would be very thanksful.

(By the way, my data is similar to yours in your n-beatsx paper.

Thank you for your time

kdgutier commented 2 years ago

Hi @linotto, we recently published a paper on the long-horizon forecasting topic

And released the implementation of a new method built on top of N-BEATS that we called N-HiTS. The paper is here: https://arxiv.org/abs/2201.12886 The code is here: https://github.com/Nixtla/neuralforecast

We have not yet tested the N-HiTSx code, but in principle one could just try it now. We are interested in your experience with it.

Greetings