First Simple Pipeline

Unit ID: ML-M03-U07 Estimated active time: 20-30 minutes

Why this matters

As projects grow, modelling code can become messy.

One cell transforms the data. Another cell fits a model. Another cell repeats the transformation on test data. It becomes easy to make mistakes.

A pipeline helps keep steps together.

The idea

A scikit-learn pipeline links steps in order.

For this first module, the pipeline is simple because we use numeric features only:

from sklearn.pipeline import Pipeline

model_pipeline = Pipeline(
    steps=[
        ("model", LogisticRegression(max_iter=1000)),
    ]
)

This is not impressive yet. It becomes more useful when preprocessing is added later.

Predict

Why introduce a pipeline before we need heavy preprocessing?

Because it builds the habit of keeping the modelling workflow together.

Run or inspect

The pipeline has the same pattern:

model_pipeline.fit(X_train, y_train)
predictions = model_pipeline.predict(X_test)

The learner does not need to master all pipeline details here. The goal is to recognise that a pipeline can behave like a model.

Change one thing

In Module 7, the pipeline will include numeric scaling and categorical encoding. That is when pipelines become essential.

For now, treat this as a preview.

Practice

Answer:

• What does a pipeline keep together?
• Why is that useful for reproducibility?
• Why are we not adding categorical preprocessing yet?

Check and explain

Complete:

A pipeline helps because the same fitted workflow can be used for ______ and ______.

Takeaway

A pipeline keeps the workflow together. Module 3 introduces the habit; Module 7 makes it necessary.