import unittest
import importlib
import numpy as np
from sklearn.linear_model import Ridge, Lasso
from sklearn.datasets import load_diabetes
from sklearn.model_selection import train_test_split
from sklearn.metrics import mean_squared_error, r2_score

def _dynamic_test(test_case, condition, success_msg, failure_msg):
    """Helper for dynamic, clear feedback."""
    if condition:
        test_case._testMethodName = success_msg
        test_case.assertTrue(True, success_msg)
    else:
        test_case._testMethodName = failure_msg
        test_case.fail(failure_msg)

class TestRidgeLassoComparison(unittest.TestCase):

    @classmethod
    def setUpClass(cls):
        """Prepare dataset and baseline models."""
        X, y = load_diabetes(return_X_y=True)
        cls.X_train, cls.X_test, cls.y_train, cls.y_test = train_test_split(
            X, y, test_size=0.3, random_state=42
        )

        # Slightly stronger regularization for clearer difference
        cls.ridge = Ridge(alpha=1.0, random_state=42).fit(cls.X_train, cls.y_train)
        cls.lasso = Lasso(alpha=0.5, random_state=42, max_iter=5000).fit(cls.X_train, cls.y_train)

    def test_r2_scores_reasonable(self):
        """Check both models produce reasonable R² scores."""
        ridge_r2 = r2_score(self.y_test, self.ridge.predict(self.X_test))
        lasso_r2 = r2_score(self.y_test, self.lasso.predict(self.X_test))
        condition = ridge_r2 > 0.35 and lasso_r2 > 0.25
        _dynamic_test(
            self,
            condition,
            "Both Ridge and Lasso achieve acceptable R² scores.",
            f"Unexpectedly low R² values: Ridge={ridge_r2:.3f}, Lasso={lasso_r2:.3f}."
        )

    def test_lasso_regularization_effect(self):
        """Check that Lasso applies stronger regularization (fewer significant coefficients)."""
        ridge_nonzero = np.sum(np.abs(self.ridge.coef_) > 1e-4)
        lasso_nonzero = np.sum(np.abs(self.lasso.coef_) > 1e-4)
        condition = lasso_nonzero <= ridge_nonzero
        _dynamic_test(
            self,
            condition,
            "Lasso correctly applies stronger regularization (fewer active coefficients).",
            f"Lasso retained too many nonzero coefficients ({lasso_nonzero}) vs Ridge ({ridge_nonzero})."
        )

    def test_mse_similarity(self):
        """Ensure Ridge and Lasso MSE values are reasonably close."""
        ridge_mse = mean_squared_error(self.y_test, self.ridge.predict(self.X_test))
        lasso_mse = mean_squared_error(self.y_test, self.lasso.predict(self.X_test))
        condition = abs(ridge_mse - lasso_mse) < 700
        _dynamic_test(
            self,
            condition,
            "Ridge and Lasso have comparable MSE values.",
            f"Large MSE difference detected: Ridge={ridge_mse:.2f}, Lasso={lasso_mse:.2f}."
        )

if __name__ == "__main__":
    unittest.main(argv=['first-arg-is-ignored'], exit=False)


test_code.py

A comprehensive intermediate course on regularization and feature selection in Python using scikit-learn, numpy, pandas, matplotlib, and seaborn. Learn to combat overfitting, apply L1/L2/Elastic Net regularization, select features using various strategies, and build robust pipelines for real-world data science tasks.

Explore the motivation for regularization, compare L1 and L2 penalties, and implement Ridge and Lasso regression.

Learn and apply filter, wrapper, and embedded feature selection methods, and integrate them into pipelines.

Delve into Elastic Net, hyperparameter tuning, coefficient visualization, and practical workflow challenges.

Challenge: Compare Ridge and Lasso on Real Data

Løsning