Skip to content

Feature importance

The importance of predictors can be obtained using the methods get_coef and get_feature_importance. These methods access the attributes coef_ and feature_importances_ of the internal regressor, respectively.

⚠ WARNING:
This methods only return values if the regressor used inside the forecaster has the attribute coef_ or feature_importances_.

Libraries

1
2
3
4
5
6
import numpy as np
import pandas as pd
from sklearn.ensemble import RandomForestRegressor
from sklearn.linear_model import Ridge
from skforecast.ForecasterAutoreg import ForecasterAutoreg
from skforecast.ForecasterAutoregMultiOutput import ForecasterAutoregMultiOutput

Data

1
2
3
4
5
6
url = ('https://raw.githubusercontent.com/JoaquinAmatRodrigo/skforecast/master/data/h2o_exog.csv')
data = pd.read_csv(url, sep=',', header=0, names=['date', 'y', 'exog_1', 'exog_2'])

data['date'] = pd.to_datetime(data['date'], format='%Y/%m/%d')
data = data.set_index('date')
data = data.asfreq('MS')

Extract feature importance from trained forecaster

1
2
3
4
5
6
7
8
forecaster = ForecasterAutoreg(
                regressor = RandomForestRegressor(),
                lags = 5
             )

forecaster.fit(y=data['y'], exog=data[['exog_1', 'exog_2']])

forecaster.get_feature_importance()
feature importance
0 lag_1 0.522544
1 lag_2 0.0998894
2 lag_3 0.0203564
3 lag_4 0.078274
4 lag_5 0.067372
5 exog_1 0.0483118
6 exog_2 0.163252 
1
2
3
4
5
6
7
8
orecaster = ForecasterAutoreg(
                    regressor = Ridge(),
                    lags = 5
                )

forecaster.fit(y=data['y'], exog=data[['exog_1', 'exog_2']])

forecaster.get_coef()
feature coef
0 lag_1 0.327688
1 lag_2 -0.0735932
2 lag_3 -0.152202
3 lag_4 -0.217106
4 lag_5 -0.1458
5 exog_1 0.379798
6 exog_2 0.668162

When using a ForecasterAutoregMultiOutput, since a different model is fit for each step, it is needed to indicate from which model extract the information.

1
2
3
4
5
6
7
8
9
forecaster = ForecasterAutoregMultiOutput(
                regressor = RandomForestRegressor(),
                steps = 10,
                lags = 5
             )

forecaster.fit(y=data['y'], exog=data[['exog_1', 'exog_2']])

forecaster.get_feature_importance(step=1)
feature importance
0 lag_1 0.529107
1 lag_2 0.109431
2 lag_3 0.0197598
3 lag_4 0.0813373
4 lag_5 0.0480799
5 exog_1 0.0371122
6 exog_2 0.175173