Closed jmrichardson closed 2 years ago
Thanks. I think I may have fixed that. Can you try
pip install --ugprade timemachines
and see if it still fails?
Oh wait my bad. Yeah it shouldn't pass e anymore. I'll fix it
Remind me where that code is?
Thank you for the fast reply! :)
Here is the location:
https://github.com/microprediction/timemachines/tree/main/examples/tuning
I've pushed a fix
Try it now? New version should be on pypi in a few minutes or you can do
pip install --upgrade git+https://github.com/microprediction/timemachines.git
By the way if you encounter more bugs feel free to let us know in the Slack, which I monitor better than issues https://join.slack.com/t/microprediction/shared_invite/zt-xauc5r03-XmIMdy~iL4a3Hw3G1Agv6w
By the way (2) you could use tuneta to create better stacking of timeseries models I suspect, since we want a collection of those whose performance is uncorrelated with each other. (i.e. it might outperform the elo_skaters)
It works now! Thanks for the fix!
Thank you for the tuneta suggestion, yes, it does make sense to include timeseries models as features. I am trying to educate myself on TS and have been using sktime. I found a reference to timemachines and wanted to try it out as it looks much more user friendly with a unified interface. I will join the slack and perhaps learn along the way. I am a little confused as to how to use in my particular use case and was hoping to get some guidance.
Hi,
I am testing the examples and get the following error with this example code: