Feat/static covs regression

[eliane-maalouf]

Fixes #1021.

Summary

• adds static covariates to the features table generated both for fit() and predict().
• It accomodates for the cases where the static covariates have different dimensions between series in a sequence and to the case where the training series might contain static covs but not the series being used for prediction

Other Information

to test outputs of the function :

import numpy as np
import pandas as pd
from darts import TimeSeries
from sklearn.preprocessing import MinMaxScaler, OneHotEncoder
from darts.dataprocessing.transformers import StaticCovariatesTransformer
from darts.utils import timeseries_generation as tg
from darts.models.forecasting.regression_model import RegressionModel

series = tg.linear_timeseries(length=10)

scaler = StaticCovariatesTransformer(transformer_cat=OneHotEncoder())

static_covs1 = pd.DataFrame(
        data={
            "cont1": [0, 1, 2],
            "cat1": [1, 2, 3],
            "cont2": [0.1, 0.2, 0.3],
            "cat2": ["a", "b", "c"],
        }
    ).astype(dtype={"cat1": "O", "cat2": "O"})
series1 = TimeSeries.from_times_and_values(
        times=series.time_index,
        values=np.concatenate([series.values()] * 3, axis=1),
        columns=["comp1", "comp2", "comp3"],
        static_covariates=static_covs1,
    )
series3 = TimeSeries.from_times_and_values(
        times=series.time_index,
        values=np.concatenate([series.values()*100] * 3, axis=1),
        columns=["comp1", "comp2", "comp3"]
    ) # series without static covs

reg = RegressionModel(lags=3, output_chunk_length=1) 

transformed_series1 = scaler.fit_transform(series1) # one hot encoding of categorical static covs

# check features (dataframe for better visibility) : 
pd.DataFrame(reg._create_lagged_data([transformed_series1, series3], None, None, max_samples_per_ts=5)[0])

[codecov-commenter]

Codecov Report

Base: 93.79% // Head: 93.79% // Increases project coverage by +0.00% 🎉

Coverage data is based on head (2c48c87) compared to base (3e5b5af).
Patch coverage: 96.77% of modified lines in pull request are covered.

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #1412   +/-   ##
=======================================
  Coverage   93.79%   93.79%           
=======================================
  Files          94       94           
  Lines        9408     9425   +17     
=======================================
+ Hits         8824     8840   +16     
- Misses        584      585    +1     

Impacted Files	Coverage Δ
darts/utils/data/inference_dataset.py	94.73% <ø> (ø)
darts/models/forecasting/regression_model.py	97.23% <83.33%> (-0.31%)	⬇️
darts/utils/data/tabularization.py	100.00% <100.00%> (ø)
darts/timeseries.py	91.85% <0.00%> (-0.06%)	⬇️
...arts/models/forecasting/torch_forecasting_model.py	89.50% <0.00%> (-0.05%)	⬇️
darts/models/forecasting/block_rnn_model.py	98.24% <0.00%> (-0.04%)	⬇️
darts/models/forecasting/nhits.py	99.27% <0.00%> (-0.01%)	⬇️
darts/datasets/__init__.py	100.00% <0.00%> (ø)

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

☔ View full report at Codecov.
📢 Do you have feedback about the report comment? Let us know in this issue.

[hrzn]

LGTM ! We can merge once the new test is added (and also maybe after we move the new function into tabularization.py).

[hrzn]

LGTM, nice one @eliane-maalouf !