SciPy with scikit-learn pipeline - Step-by-Step Execution

Concept Flow - scikit-learn pipeline

Load Data

↓

Define Pipeline Steps

↓

Create Pipeline Object

↓

Fit Pipeline on Training Data

↓

Predict or Transform Data

↓

Evaluate or Use Results

This flow shows how data is loaded, a pipeline is created with steps, then fitted and used for prediction or transformation.

Execution Sample

SciPy

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

pipe = Pipeline([
  ('scaler', StandardScaler()),
  ('logreg', LogisticRegression())
])

pipe.fit(X_train, y_train)
preds = pipe.predict(X_test)

This code creates a pipeline that scales data then fits a logistic regression model, then predicts on test data.

Execution Table

Step	Action	Input Data Shape	Output Data Shape	Notes
1	Load X_train, y_train	(100, 4), (100,)	(100, 4), (100,)	Data loaded with 100 samples, 4 features
2	Create Pipeline with scaler and logistic regression	N/A	Pipeline object created	Pipeline ready with two steps
3	Fit pipeline on X_train, y_train	(100, 4), (100,)	Model fitted	Scaler fit and applied, logistic regression trained
4	Predict on X_test	(20, 4)	(20,)	Predictions generated for 20 test samples
5	Output predictions	(20,)	(20,)	Final predicted labels array
6	End	N/A	N/A	Pipeline execution complete

💡 All steps completed; pipeline fit and predictions done

Variable Tracker

Variable	Start	After Step 1	After Step 3	After Step 4	Final
X_train	undefined	(100, 4)	(100, 4)	(100, 4)	(100, 4)
y_train	undefined	(100,)	(100,)	(100,)	(100,)
pipe	undefined	Pipeline object	Fitted pipeline	Fitted pipeline	Fitted pipeline
preds	undefined	undefined	undefined	(20,)	(20,)

Key Moments - 3 Insights

Why do we fit the pipeline instead of fitting scaler and model separately?

Fitting the pipeline (see execution_table step 3) ensures the scaler is fit only on training data and the same scaling is applied during prediction, avoiding data leakage.

What shape does the data have after scaling inside the pipeline?

The shape remains the same (100 samples, 4 features) as scaling changes values but not dimensions, shown in variable_tracker for X_train after step 3.

Why do predictions have shape (20,) after step 4?

Because predictions are labels for each test sample, one per sample, so 20 samples produce 20 predicted labels, as shown in execution_table step 4.

Visual Quiz - 3 Questions

Look at the execution_table at step 3, what happens during pipeline fitting?

A Scaler is fit on test data

B Only logistic regression is fit, scaler is ignored

C Scaler and logistic regression are both fit on training data

D Pipeline predicts without fitting

💡 Hint

Refer to execution_table row with Step 3 describing fitting process

According to variable_tracker, what is the shape of preds after step 4?

A (20,)

B (100, 4)

C (100,)

D (20, 4)

💡 Hint

Check preds variable column 'After Step 4' in variable_tracker

If we skip scaling in the pipeline, how would execution_table step 3 change?

A Pipeline would fail to fit

B Scaler fit step would be missing, model fit remains same

C Predictions would be shape (100, 4)

D Data shape would change to (20,)

💡 Hint

Think about what happens if scaler step is removed from pipeline steps

Concept Snapshot

scikit-learn pipeline:
- Use Pipeline to chain steps like scaling and modeling
- Fit pipeline on training data to avoid data leakage
- Predict or transform data using pipeline
- Keeps code clean and reproducible
- Data shape stays consistent through pipeline steps

Full Transcript

This visual execution shows how to use a scikit-learn pipeline. First, data is loaded with 100 samples and 4 features. Then a pipeline is created with two steps: StandardScaler and LogisticRegression. The pipeline is fit on training data, which fits the scaler and model in order. After fitting, predictions are made on test data with 20 samples. Variables like X_train, y_train, pipeline object, and predictions change state through the steps. Key moments clarify why fitting the pipeline is important to avoid data leakage, how data shape remains the same after scaling, and why predictions have shape matching test samples. The quiz tests understanding of pipeline fitting, prediction shapes, and effects of skipping scaling. The snapshot summarizes pipeline usage for clean, reproducible modeling.