search

AIStream-Peelout / flow-forecast

Deep learning PyTorch library for time series forecasting, classification, and anomaly detection (originally for flood forecasting).
https://flow-forecast.atlassian.net/wiki/spaces/FF/overview
GNU General Public License v3.0
1.97k stars 288 forks source link

How to use multiple discrete time series for model training #688

Open taylornelsonbluecargo opened 11 months ago

taylornelsonbluecargo commented 11 months ago

Hi,

I'm looking to train on multiple datasets which are temporal but all start at time 0 and proceed for some amount of time. I can't stack these time series together because each is a different event to conflate them would ruin the predictive capability of any model afaik. I want to train across several thousand samples of variable time length. I'm not sure how to do this with flow forecast or if you could point me in the right direction that would be helpful.

This is analogous to training on a subset of stock data. I have 10 stocks for 1 day time series each should have discrete training as they have underlying variables which differ. Any advice would be helpful happy to share the csvs I am working with. Thanks!

isaacmg commented 11 months ago

Could you give a sample of the data? You could train a model for each, concat them all together/do multi targets or you could use a series id that you pass in for each. If I'm understanding your question. properly.