Pipelines in Python

A pipeline chains preprocessing and a model into one estimator, so nothing leaks.

from sklearn.linear_model import LogisticRegression
from sklearn.pipeline import Pipeline, make_pipeline
from sklearn.preprocessing import StandardScaler

named = Pipeline([
    ("scale", StandardScaler()),
    ("model", LogisticRegression(max_iter=200)),
])

quick = make_pipeline(StandardScaler(), LogisticRegression())

X = [[0.0, 1.0], [1.0, 0.5], [2.0, 0.0], [3.0, 0.5]]
y = [0, 0, 1, 1]

named.fit(X, y)
print(named.predict([[1.5, 0.25]]))
print(named.named_steps["scale"].mean_)
print(quick.fit(X, y).score(X, y))

How it works

1. Each step is a (name, estimator) pair; the last one predicts.
2. fit fits every step in turn on the training data only.
3. make_pipeline names the steps after their classes.

Keywords and builtins used here

• print

The run, in numbers

Lines

18

Characters to type

512

Tokens

154

Three-star pace

105 tpm

At the three-star pace of 105 tokens a minute, this run takes about 88 seconds.

Type this snippet

Step 4 of 4 in Preprocessing, step 7 of 25 in Machine learning with scikit-learn.

← Previous Next →

Pipelines in other languages

• Go