Generally, a machine learning pipeline is a series of steps, executed in an order wise to automate the machine learning workflows. A series of steps include training, splitting, and deploying the model.
Pipeline
It is used to execute the process sequentially and execute the steps, transformers, or estimators are named manually. Transformers and estimators are the parameters to fit the model and tune for its model accuracy.
Syntax: class sklearn.pipeline.Pipeline(steps, *, memory=None, verbose=False)
Note: In the above syntax, steps(can be represented as an array) are represented as a sequence of functions to be executed in completing in the task.
Example:
Here we are going to create a pipeline using Pipeline() methods.
Python3
# import required modules import numpy as np from sklearn.svm import SVC from sklearn.preprocessing import StandardScaler from sklearn.pipeline import Pipeline # declare X, used as a feature with # nested array X = np.array([[1, 2], [3, 4], [5, 6], [7, 8]]) # declare y as target variable y = np.array([9, 10, 11, 12]) # pipeline is created using Pipeline() pipe = Pipeline([('std_sc', StandardScaler()), ('svc_', SVC(gamma='auto'))]) pipe.fit(X, y) |
Output:
make_pipeline
make_pipleine is an advanced method in scikit learn, in which the naming of the estimators or transformers are done automatically.
Syntax: sklearn.pipeline.make_pipeline(*steps, memory=None, verbose=False)
Example:
Here we are going to make pipeline using make_pipeline() methods.
Python3
# import required modules import numpy as np from sklearn.pipeline import make_pipeline from sklearn.preprocessing import StandardScaler from sklearn.svm import SVC # declare X, used as a feature with # nested array X = np.array([[1, 2], [3, 4], [5, 6], [7, 8]]) # declare y as target variable y = np.array([9, 10, 11, 12]) # pipeline is created using make_pipeline() mp = make_pipeline(StandardScaler(), SVC(gamma='auto')) mp.fit(X, y) |
Output:
Table of difference between pipeline and make_pipeline in scikit
|
pipeline |
make_pipeline |
|---|---|
| The pipeline requires naming the steps, manually. | make_pipeline names the steps, automatically. |
| Names are defined explicitly, without rules. | Names are generated automatically using a straightforward rule (lower case of the estimator). |
| Names cannot be changed based on the transformer or estimator used. | Names are readable, short, and easy to understand, and can be changed based on the estimator used. |
