Can scikit-learn's LogisticRegression() automatically normalize input data to z-scores?

Question

Is there a way to have an instance of LogisticRegression() automatically normalize the data supplied for fitting/training to z-scores to build the model? LinearRegression() has a normalize=True parameter but maybe this doesn't make sense for LogisticRegression()?

If so, would I have to normalize unlabeled input vectors by hand (i.e., recalculate the mean, standard deviation for each column) before calling predict_proba()? This would be strange if the model already performed that possibly costly computation.

Thanks

Answer 1

Is this what you are looking for?

from sklearn.datasets import make_classification
from sklearn.preprocessing import StandardScaler
from sklearn.pipeline import make_pipeline
from sklearn.linear_model import LogisticRegression


X, y = make_classification(n_samples=1000, n_features=100, weights=[0.1, 0.9], random_state=0)
X.shape

# build pipe: first standardize by substracting mean and dividing std
# next do classificaiton
pipe = make_pipeline(StandardScaler(), LogisticRegression(class_weight='auto'))

# fit
pipe.fit(X, y)
# predict
pipe.predict_proba(X)

# to get back mean/std
scaler = pipe.steps[0][1]
scaler.mean_
Out[12]: array([ 0.0313, -0.0334,  0.0145, ..., -0.0247,  0.0191,  0.0439])

scaler.std_
Out[13]: array([ 1.    ,  1.0553,  0.9805, ...,  1.0033,  1.0097,  0.9884])