Added Ridge Regression To Machine Learning #12108

machine_learning/ridge_regression.py

@@ -0,0 +1,104 @@
+"""
+Ridge Regression with L2 Regularization using Gradient Descent.
+
+Ridge Regression is a type of linear regression that includes an L2 regularization
+term to prevent overfitting and improve generalization. It is commonly used when
+multicollinearity is present in the data.
+
+More on Ridge Regression: https://en.wikipedia.org/wiki/Tikhonov_regularization
+"""
+
+from typing import Tuple
+import numpy as np
+import pandas as pd
+
+def load_data(file_path: str) -> Tuple[np.ndarray, np.ndarray]:
+    """
+    Load data from a CSV file and return features and target arrays.
+
+    Args:
+        file_path: Path to the CSV file.
+
+    Returns:
+        A tuple containing features (X) and target (y) as numpy arrays.
+
+    Example:
+    >>> data = pd.DataFrame({'ADR': [200, 220], 'Rating': [1.2, 1.4]})
+    >>> data.to_csv('sample.csv', index=False)
+    >>> X, y = load_data('sample.csv')
+    >>> X.shape == (2, 1) and y.shape == (2,)
+    True
+    """
+    data = pd.read_csv(file_path)
+    X = data[['Rating']].to_numpy()  # Use .to_numpy() instead of .values (PD011)
+    y = data['ADR'].to_numpy()
+    return X, y
+
+def ridge_gradient_descent(
+    X: np.ndarray, y: np.ndarray, reg_lambda: float, learning_rate: float,
+    num_iters: int = 1000
+) -> np.ndarray:
+    """
+    Perform Ridge Regression using gradient descent.
+
+    Args:
+        X: Feature matrix.
+        y: Target vector.
+        reg_lambda: Regularization parameter (lambda).
+        learning_rate: Learning rate for gradient descent.
+        num_iters: Number of iterations for gradient descent.
+
+    Returns:
+        Optimized weights (coefficients) for predicting ADR from Rating.
+
+    Example:
+    >>> X = np.array([[1.2], [1.4]])
+    >>> y = np.array([200, 220])
+    >>> ridge_gradient_descent(X, y, reg_lambda=0.1, learning_rate=0.01).shape == (1,)
+    True
+    """
+    weights = np.zeros(X.shape[1])
+    m = len(y)
+
+    for _ in range(num_iters):
+        predictions = X @ weights
+        error = predictions - y
+        gradient = (X.T @ error + reg_lambda * weights) / m
+        weights -= learning_rate * gradient
+
+    return weights
+
+def mean_absolute_error(y_true: np.ndarray, y_pred: np.ndarray) -> float:
+    """
+    Calculate the Mean Absolute Error (MAE) between true and predicted values.
+
+    Args:
+        y_true: Actual values.
+        y_pred: Predicted values.
+
+    Returns:
+        Mean absolute error.
+
+    Example:
+    >>> mean_absolute_error(np.array([200, 220]), np.array([205, 215]))
+    5.0
+    """
+    return np.mean(np.abs(y_true - y_pred))
+
+if __name__ == "__main__":
+    import doctest
+    doctest.testmod()
+
+    # Load the data
+    X, y = load_data("sample.csv")
+
+    # Fit the Ridge Regression model
+    optimized_weights = ridge_gradient_descent(X, y, reg_lambda=0.1, learning_rate=0.01)
+
+    # Make predictions
+    y_pred = X @ optimized_weights
+
+    # Calculate Mean Absolute Error
+    mae = mean_absolute_error(y, y_pred)
+    print("Optimized Weights:", optimized_weights)
+    print("Mean Absolute Error:", mae)