search

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
1.02k stars 55 forks source link

_fit() method throws ShapeError #157

Open Alla-Z opened 9 months ago

Alla-Z commented 9 months ago

Describe the bug The quick start code throws an exception when run on my panel data frame, specifically in the _fit() method. The data frame schema is : OrderedDict([('restaurant_id', UInt64), ('insertdate', Datetime(time_unit='ns', time_zone=None)), ('prep_time', Int64)]) The data frame shape is: (735_109,3)

The data frame time column is not regularly sampled therefore we ran the resampling method:

trans = resample(freq='1m', agg_method='mean', impute_method='ffill')
y= final_df.pipe(trans).collect()

We tried the forward fill. backward fill, and zeros as an impute method.

Screenshots image

Desktop:

topher-lo commented 9 months ago

Potentially related to https://github.com/functime-org/functime/issues/17

Need to create a MRE and double check whether forecast unit tests are still passing.

baggiponte commented 3 months ago

Ciao @Alla-Z, hope you are doing alright. Do you think you can manage to create a minimal reproducible example (MRE) to see whether the issue is still present in the latest versions of functime?