search

timeseriesAI / tsai

Time series Timeseries Deep Learning Machine Learning Python Pytorch fastai | State-of-the-art Deep Learning library for Time Series and Sequences in Pytorch / fastai
https://timeseriesai.github.io/tsai/
Apache License 2.0
4.92k stars 622 forks source link

Implementing PatchTST but on a different type of supervised data #825

Open aatmanvaidya opened 10 months ago

aatmanvaidya commented 10 months ago

Hi @oguiza , thank you so much for integrating PatchTST to tsai and for your helpful tutorial in the Google Colab. But I have a bit of different data, and I need some help in applying the patchTST model from tsai in it.

I have a time series Twitter dataset that looks like this. The first column is the time column, and the rest are the user id. Each row of the user ID column has a toxicity value for that time. The dataset looks like this time 378843212 1246821236 32186559 65458298 19017659
2017-01-01 00:00:03 0.041532 0.038335 0.047807 0.050232 0.034241
2017-01-01 01:00:03 0.041532 0.038335 0.047807 0.050232 0.034241
2017-01-01 02:00:03 0.041532 0.038335 0.047807 0.050232 0.034241
2017-01-01 03:00:03 0.041532 0.038335 0.047807 0.050232 0.034241
2017-01-01 04:00:03 0.041532 0.038335 0.047807 0.050232 0.034241

So now, using the patchTST model, I want to forecast this time series for each user, so my model should be able to tell a toxicity value for a user at a certain time. How can I do this? It would be of great help to me if you can help me out?

oguiza commented 10 months ago

Hi @aatmanvaidya, this is not yet currently supported by PatchTST. For now, you can only use it with univariate datasets, or multivariate datasets where you predict all input variables (in fact, this is duplicate of #713).

aatmanvaidya commented 10 months ago

Hello @oguiza, understood, thank you so much for your reply. Apologies, I missed out that a similar issue already existed.