End-to-end modeling
Combine preprocessing and prediction into one fit/predict object.
Scikit-Learn · Function
make_pipeline
Constructs a scikit-learn Pipeline from estimators supplied in order. It automatically generates lowercase step names from estimator types.
How it works
Pass transformers and a final estimator in execution order. The function builds a Pipeline, automatically deriving each step name by lowercasing its estimator type; the resulting pipeline can be fitted and used for prediction as a single estimator.
Syntax
sklearn.pipeline.make_pipeline(*steps, memory=None, transform_input=None, verbose=False)Parameters
| Name | Type | Description |
|---|---|---|
steps | estimators | Ordered estimator objects to chain into the pipeline. |
memory | str or object, default=None | Cache location or a joblib.Memory-compatible object; default=None. |
transform_input | list of str, default=None | Names of fit inputs other than X to transform through relevant preceding steps; default=None. |
verbose | bool, default=False | Whether to print elapsed fitting time for each step; default=False. |
Returns
A scikit-learn Pipeline object.
Key points
- The function creates a Pipeline from the estimators passed as positional steps.
- Step names are generated automatically by lowercasing each estimator's type name.
- The positional steps form an ordered chain of scikit-learn estimators.
- The memory option can cache fitted transformer steps.
- verbose=True prints elapsed fitting time after each step completes.
Common mistakes
- Enabling memory caching does not cache the final pipeline step.
- With memory caching enabled, inspect the fitted clones through the pipeline rather than relying on the original transformer objects.
Example: Scale features and classify
from sklearn.datasets import load_iris
from sklearn.linear_model import LogisticRegression
from sklearn.pipeline import make_pipeline
from sklearn.preprocessing import StandardScaler
X, y = load_iris(return_X_y=True)
model = make_pipeline(StandardScaler(), LogisticRegression(max_iter=200))
model.fit(X, y)
print(model.predict(X[:3]))Standardizes the features and then fits a logistic-regression classifier as one estimator.
When to use it
Consistent preprocessing
Reuse the same transformation sequence during validation and inference.
Transformer caching
Avoid repeating expensive transformer fitting across compatible runs.
Interview questions
QWhat does make_pipeline provide compared with Pipeline?
make_pipeline accepts a variable number of estimators and returns a Pipeline containing them in the supplied order. It automatically names each step from the lowercase estimator type, so explicit step names are not provided.
Follow-up: When are explicit names preferable? Use Pipeline when stable, custom names are needed for parameter searches or inspection.
QHow are the supplied estimators used?
The steps are chained sequentially: transformers process the data before the final estimator receives it. The returned Pipeline exposes the composite estimator interface, including fitting and prediction when its last step supports those operations.
Follow-up: How would you inspect a particular step? Access the pipeline's named_steps or steps attributes.
QWhat does the memory parameter do?
memory accepts a path or an object implementing the joblib.Memory interface. It caches fitted transformers, while caching is disabled by default and the final step is excluded from caching.
Follow-up: What additional behavior should be considered when caching? Transformer instances are cloned before fitting, so inspect the fitted pipeline steps.
Related
Pipelinemake_unionFeatureUnion