train_models.py

from sklearn.metrics import mean_squared_error
from sklearn.model_selection import train_test_split
from sklearn.linear_model import Lasso, Ridge, LinearRegression,ElasticNet
from sklearn.model_selection import cross_val_score

def train_eval_model(train_housing_prepared, train_labels, strategy, alpha = None, l1_ratio = None):
    if alpha != None:
        print('alpha: ', alpha)
        if l1_ratio != None:
            print('l1_ratio: ', l1_ratio)
            lin_reg = strategy(alpha = alpha, l1_ratio = l1_ratio, max_iter=100000) #LinearRegression()
        else:
            lin_reg = strategy(alpha = alpha, max_iter=50000)
    else:
        lin_reg = strategy()
    
    lin_reg.fit(train_housing_prepared, train_labels)
    #print('coefs: ', lin_reg.coef_)

    housing_predictions = lin_reg.predict(train_housing_prepared)
    train_mse = mean_squared_error(housing_predictions, train_labels)
    train_rmse = np.sqrt(train_mse)
    print('train rmse: ', train_rmse)

    test_housing_predictions = lin_reg.predict(test_housing_prepared)
    test_mse = mean_squared_error(test_housing_predictions, test_labels)
    test_rmse = np.sqrt(test_mse)
    print('test rmse: ', test_rmse)
    
    
    return test_rmse

train_set, test_set = train_test_split(housing.loc[housing_labels_filtered.index], test_size=0.2, random_state=42)

train_housing_prepared = full_pipeline.fit_transform(train_set)
test_housing_prepared = full_pipeline.transform(test_set)

train_labels = train_set['median_house_value']
test_labels = test_set['median_house_value']


cute_results = '***** \n'

for strategy, strategy_name in [(Lasso, 'Lasso'), (Ridge, 'Ridge')]:
    print('---*---*---')
    print(strategy_name)
    min_test_rmse = float('inf')
    results = ''
    for alpha in [.1, .5, 1, 5, 10, 50, 100, 500, 1000, 5000, 10000, 15000]:
        new_test_rmse = train_eval_model(train_housing_prepared, train_labels, strategy, alpha)
        if new_test_rmse < min_test_rmse :
            min_test_rmse = new_test_rmse
            results = str([strategy_name, ': rmse: ' , min_test_rmse, 'alpha: ', alpha])
    cute_results+=results
    cute_results+='\n'
print('---*---*---')
for strategy, strategy_name in [(ElasticNet, 'ElasticNet')]:
    print(strategy_name)
    results = ''
    min_test_rmse = float('inf')
    for alpha in [.1, .5, 1, 5, 10, 50, 100, 500, 1000, 5000, 10000, 15000]:
        for l1_ratio in [.1, .25, .5, .75, .9]:
            new_test_rmse = train_eval_model(train_housing_prepared, train_labels, strategy, alpha, l1_ratio)
            if new_test_rmse < min_test_rmse :
                min_test_rmse = new_test_rmse
                results = str([strategy_name, ': rmse: ' , min_test_rmse, 'alpha: ', alpha, 'l1_ratio:', l1_ratio])
    cute_results+=results
    cute_results+='\n'
    
print('---*---*---')
results = train_eval_model(train_housing_prepared, train_labels, LinearRegression)
cute_results += str(['linearRegression: rmse: ', results])