functime-org / functime

Time-series machine learning at scale. Built with Polars for embarrassingly parallel feature extraction and forecasts on panel data.
https://docs.functime.ai
Apache License 2.0
971 stars 52 forks source link

Add float support in `train_test_split` #168

Closed FBruzzesi closed 4 months ago

FBruzzesi commented 6 months ago

Description

This is a first draft of how I would address and fix #167 to add support for float in train_test_split. I noticed typing has some issues but it is already been addresses in #161.

Question: Would you like to have a default value (e.g. 25% as in scikit-learn) for the test size?

vercel[bot] commented 6 months ago

The latest updates on your projects. Learn more about Vercel for Git ↗︎

Name Status Preview Comments Updated (UTC)
functime-docs ✅ Ready (Inspect) Visit Preview 💬 Add feedback Jan 12, 2024 3:06pm
baggiponte commented 6 months ago

Will check later today - for the meantime, thanks a lot! Wanted to get this done. Would merge #161 before this. Any comments on that?

FBruzzesi commented 6 months ago

Thanks for getting back to me, let me know what you think regarding having a default value 😊 After that I would say that I can mark it as ready for review

baggiponte commented 6 months ago

Thanks for getting back to me, let me know what you think regarding having a default value 😊 After that I would say that I can mark it as ready for review

I would like to have a default value. We might go with sklearn's default. You worked in the field though, what would you use as a reasonable value? @ngriffiths13 might also know some heuristics.

ngriffiths13 commented 5 months ago

Thanks for getting back to me, let me know what you think regarding having a default value 😊 After that I would say that I can mark it as ready for review

I would like to have a default value. We might go with sklearn's default. You worked in the field though, what would you use as a reasonable value? @ngriffiths13 might also know some heuristics.

I would use .25 as the default value.

baggiponte commented 5 months ago

Seems great! @ngriffiths13 how are we with tests, did you manage to reduce the duration? Should Francesco add the tests for the new train_test_split?

FBruzzesi commented 5 months ago

Should Francesco add the tests for the new train_test_split?

Added a few basic tests