tsflex

tsflex copied to clipboard

Suppose you want to (1) extract features for multiple groups (e.g. multiple users or machines), and (2) your data is irregularly sampled (or gaps occur), and (3) you want to...

jonasvdd

jvdd

polars [`groupby_dynamic`](https://pola-rs.github.io/polars-book/user-guide/howcani/timeseries/temporal_groupby.html#parameters-for-groupby_dynamic) seems worthwile to further explore :mag: If this seems rather performant, this might be an interesting alternative for our strided-rolling backend :eyes:

jvdd

jvdd

Do not assume datetimeindex in the processing functionality

1

jvdd

Related to https://github.com/predict-idlab/tsflex/pull/120#discussion_r1570385622

jvdd

:tada: allow non `pd.DatetimeIndex` for processing

1

Closes #125 TODO - [ ] update processing tests (currently only checks for datetimeindex)

jvdd

Stuck on calculating features

8

cam@DESKTOP-41QNH42:/mnt/c/Users/ahobb/PycharmProjects/cammm$ poetry run python cammm/feature_eng/extract.py 99%|████████████████████████████████████████████████████████████████████████████████ | 273/276 [00:19

cameron-hobbs