Sklearn components in pipeline is not fitted even if the whole pipeline is?

问题

I'm trying to single out a component/transformer from a fitted pipeline to inspect it's behavior. However, when I retrieved the component, the component is showed as unfitted, but using the pipeline as a whole works without problem. This suggest the pipeline is fitted and the components are fitted as well.

Can someone explain why, and also suggest how to inspect a component in a fitted pipeline?

Here's a reproducible example:

import pandas as pd
import numpy as np

from sklearn.compose import ColumnTransformer
from sklearn.pipeline import Pipeline
from sklearn.impute import SimpleImputer
from sklearn.preprocessing import StandardScaler, OneHotEncoder
from sklearn.linear_model import LogisticRegression
from sklearn.model_selection import train_test_split, GridSearchCV

np.random.seed(0)

# Read data from Titanic dataset.
titanic_url = ('https://raw.githubusercontent.com/amueller/'
               'scipy-2017-sklearn/091d371/notebooks/datasets/titanic3.csv')
data = pd.read_csv(titanic_url)

# We create the preprocessing pipelines for both numeric and categorical data.
numeric_features = ['age', 'fare']
numeric_transformer = Pipeline(steps=[
    ('imputer', SimpleImputer(strategy='median')),
    ('scaler', StandardScaler())])

categorical_features = ['embarked', 'sex', 'pclass']
categorical_transformer = Pipeline(steps=[
    ('imputer', SimpleImputer(strategy='constant', fill_value='missing')),
    ('onehot', OneHotEncoder(handle_unknown='ignore'))])

preprocessor = ColumnTransformer(
    transformers=[
        ('num', numeric_transformer, numeric_features),
        ('cat', categorical_transformer, categorical_features)])

# Append classifier to preprocessing pipeline.
# Now we have a full prediction pipeline.
clf = Pipeline(steps=[('preprocessor', preprocessor),
                      ('classifier', LogisticRegression(solver='lbfgs'))])

X = data.drop('survived', axis=1)
y = data['survived']

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2)

clf.fit(X_train, y_train)
print("model score: %.3f" % clf.score(X_test, y_test))

Calling either:

clf.get_params()['preprocessor__cat__imputer'].transform(X)

clf.named_steps['preprocessor'].transformers[0][1].named_steps['imputer'].transform(X)

will result in such error:

NotFittedError: This SimpleImputer instance is not fitted yet. Call 'fit' with appropriate arguments before using this method.

回答1:

The ColumnTransformer attribute transformers is the input unfitted transformers. To access the fitted transformers, use the attribute transformers_ or named_transformers_. I suppose get_params()['preprocessor__cat__imputer'] is also getting the unfitted input transformer.

(You'll still get an error, because the imputer will try to work on the string data as well, and strategy='median' will fail.)

来源：https://stackoverflow.com/questions/58704347/sklearn-components-in-pipeline-is-not-fitted-even-if-the-whole-pipeline-is

标签

python

scikit-learn

pipeline