# Select relevant features
features = ['R', 'AB', 'H', '2B', '3B', 'HR', 'BB', 'SO', 'SB', 'CS', 'HBP', 'SF', 'RA', 'ER', 'CG', 'SHO', 'SV', 'HA', 'HRA', 'BBA', 'SOA', 'E', 'DP', 'FP', 'attendance', 'BPF', 'PPF', 'differential', 'AVG', 'OBP', '1B', 'SLG', 'salary']
# Split the data into features (X) and the target variable (y)
X = baseball[features]
y = baseball['W']
# Shuffle X and y to avoid data being grouped by year
X_shuffled, y_shuffled = shuffle(X, y, random_state=42)
# Split the shuffled data into training and testing sets
X_train, X_test, y_train, y_test = train_test_split(X_shuffled, y_shuffled, test_size=0.2, random_state=42)
# Standardizing the features, a good all-purpose scaling method
scaler = StandardScaler()
X_train_scaled = scaler.fit_transform(X_train)
X_test_scaled = scaler.transform(X_test)
# Lasso Regression with cross-validation
lasso_model = Lasso(alpha=1.0)
lasso_cv_results = cross_validate(lasso_model, X_train_scaled, y_train, cv=5, scoring={'mse': make_scorer(mean_squared_error), 'mae': make_scorer(mean_absolute_error), 'mape': 'neg_mean_absolute_percentage_error'}, return_train_score=True)
# Ridge Regression with cross-validation
ridge_model = Ridge(alpha=1.0)
ridge_cv_results = cross_validate(ridge_model, X_train_scaled, y_train, cv=5, scoring={'mse': make_scorer(mean_squared_error), 'mae': make_scorer(mean_absolute_error), 'mape': 'neg_mean_absolute_percentage_error'}, return_train_score=True)
# Outputting cross-validation results
print("Lasso Regression Cross-Validation Results:")
print("Train MSE:", lasso_cv_results['train_mse'])
print("Test MSE:", lasso_cv_results['test_mse'])
print("Train MAE:", lasso_cv_results['train_mae'])
print("Test MAE:", lasso_cv_results['test_mae'])
print("Train MAPE:", -lasso_cv_results['train_mape'].mean())
print("Test MAPE:", -lasso_cv_results['test_mape'].mean())
print("\nRidge Regression Cross-Validation Results:")
print("Train MSE:", ridge_cv_results['train_mse'])
print("Test MSE:", ridge_cv_results['test_mse'])
print("Train MAE:", ridge_cv_results['train_mae'])
print("Test MAE:", ridge_cv_results['test_mae'])
print("Train MAPE:", -ridge_cv_results['train_mape'].mean())
print("Test MAPE:", -ridge_cv_results['test_mape'].mean())
# Fit final models on the entire training set
lasso_model.fit(X_train_scaled, y_train)
ridge_model.fit(X_train_scaled, y_train)
# Making predictions on the test set
y_pred_lasso = lasso_model.predict(X_test_scaled)
y_pred_ridge = ridge_model.predict(X_test_scaled)
# Evaluating performance on the test set
mse_lasso = mean_squared_error(y_test, y_pred_lasso)
mse_ridge = mean_squared_error(y_test, y_pred_ridge)
mae_lasso = mean_absolute_error(y_test, y_pred_lasso)
mae_ridge = mean_absolute_error(y_test, y_pred_ridge)
mape_lasso = mean_absolute_error(y_test, y_pred_lasso) / y_test.mean() * 100
mape_ridge = mean_absolute_error(y_test, y_pred_ridge) / y_test.mean() * 100
# Outputting additional metrics
print("\nLasso Regression Metrics on Test Set:")
print("Test MSE:", mse_lasso)
print("Test MAE:", mae_lasso)
print("Test MAPE:", mape_lasso)
print("\nRidge Regression Metrics on Test Set:")
print("Test MSE:", mse_ridge)
print("Test MAE:", mae_ridge)
print("Test MAPE:", mape_ridge)
# Calculating coefficients
linear_coefficients = pd.Series(lasso_model.coef_, index=features, name='Lasso Coefficients')
ridge_coefficients = pd.Series(ridge_model.coef_, index=features, name='Ridge Coefficients')
# Outputting the coefficients
print("\nLasso Regression Coefficients:")
print(linear_coefficients)
print("\nRidge Regression Coefficients:")
print(ridge_coefficients)
# Plotting the coefficients
plt.figure(figsize=(12, 6))
linear_coefficients.plot(kind='bar', color='blue', alpha=0.7, label='Lasso Regression')
ridge_coefficients.plot(kind='bar', color='orange', alpha=0.7, label='Ridge Regression')
plt.title('Lasso vs. Ridge Regression Coefficients')
plt.xlabel('Features')
plt.ylabel('Coefficients')
plt.legend()
plt.show()
