How to train logistic regression with scikit-learn

A classification model earns its place by predicting labels for examples it did not see during fitting. scikit-learn provides LogisticRegression as a compact linear baseline for labeled numeric data, with predicted labels and class probabilities available through the estimator API.

The Iris dataset supplies three species and four numeric flower measurements. A fixed stratified split reserves 25 percent of its rows for testing while preserving the species mix, so the fitted model is evaluated against data excluded from training.

Placing StandardScaler and LogisticRegression in the same Pipeline prevents the scaler from learning from the test partition. The completed model can then apply that fitted transformation consistently when it scores the held-out rows and predicts an unseen flower.

1. Create train_logistic_regression.py with the imports and stratified Iris train/test split.

   train_logistic_regression.py

   from sklearn.datasets import load_iris
   from sklearn.linear_model import LogisticRegression
   from sklearn.model_selection import train_test_split
   from sklearn.pipeline import make_pipeline
   from sklearn.preprocessing import StandardScaler
    
    
   iris = load_iris()
   X_train, X_test, y_train, y_test = train_test_split(
       iris.data,
       iris.target,
       test_size=0.25,
       stratify=iris.target,
       random_state=42,
   )

2. Append the scaled logistic regression pipeline after the train/test split.

   model = make_pipeline(
       StandardScaler(),
       LogisticRegression(max_iter=200),
   )
   model.fit(X_train, y_train)

   The pipeline fits StandardScaler from X_train and applies the learned transformation automatically during prediction.
   Related: How to create a scikit-learn pipeline
   Related: How to standardize features with scikit-learn

3. Append the held-out evaluation block beneath the existing pipeline section.

   accuracy = model.score(X_test, y_test)
   sample = X_test[[0]]
   prediction = model.predict(sample)[0]
   probabilities = model.predict_proba(sample)[0]
    
   assert accuracy >= 0.90
   assert abs(probabilities.sum() - 1.0) < 1e-12
    
   class_probabilities = {
       str(name): round(float(probability), 3)
       for name, probability in zip(iris.target_names, probabilities)
   }
    
   print(f"training rows: {len(X_train)}")
   print(f"test rows: {len(X_test)}")
   print(f"held-out accuracy: {accuracy:.3f}")
   print(f"predicted species: {iris.target_names[prediction]}")
   print(f"class probabilities: {class_probabilities}")
   print(f"probability sum: {probabilities.sum():.3f}")

   The assertions make the run fail if held-out accuracy drops below 0.90 or the predicted class probabilities stop forming a complete distribution.

4. Run the completed logistic regression script from the directory that contains it.

   $ python3 train_logistic_regression.py
   training rows: 112
   test rows: 38
   held-out accuracy: 0.921
   predicted species: setosa
   class probabilities: {'setosa': 0.985, 'versicolor': 0.015, 'virginica': 0.0}
   probability sum: 1.000

   The observed score comes from 38 held-out rows, while the prediction and probability output exercise the fitted pipeline on one unseen flower.