}
}
}
scores = []
models = {}
for model_name, mp in model_params.items():
clf = GridSearchCV(mp["model"],
mp["params"],
cv=6,
return_train_score=False)
clf.fit(X_train, y_train)
scores.append({
"model_name": model_name,
"best_score": clf.best_score_,
"best_params": clf.best_params_
})
models[model_name] = clf
scores_df = pd.DataFrame(
scores, columns=['model_name', 'best_score', 'best_params'])
scores_df.sort_values(by='best_score', ascending=False, inplace=True)
print(scores_df)
best_model_name = scores_df.iloc[0]['model_name']
best_clf = models[best_model_name]
model_filepath = args['model']
print(f'≫ Storing "{best_model_name}" in "{model_filepath}"')
store_model(best_clf, model_filepath)
parameters = {
'fit_intercept': [True, False],
'normalize': [True, False],
}
model = GridSearchCV(linreg, param_grid=parameters)
model.fit(X_train, y_train)
print('> Best parameters:', model.best_params_)
y_pred = model.predict(X_test)
r2 = evaluate_regression(y_test, y_pred)
models[r2] = model.best_estimator_
print('Random Forest')
forest = RandomForestRegressor()
parameters = {
'n_estimators': [75, 100, 300],
'max_depth': [2, 5, None],
}
model = GridSearchCV(forest, param_grid=parameters)
model.fit(X_train, y_train)
print('> Best parameters:', model.best_params_)
y_pred = model.predict(X_test)
r2 = evaluate_regression(y_test, y_pred)
models[r2] = model.best_estimator_
best = models[max(models)]
model_filepath = args['model']
print(f'≫ Storing Model {type(best)} in "{model_filepath}"')
store_model(best, model_filepath)
def test_filepath_format():
with pytest.raises(ValueError, match="filepath should end with specific extension"):
store_model(model, 'filepath')
def test_new_number_of_columns(dump_mock):
store_model(model, 'filepath.pkl')
dump_mock.assert_called_once_with(model, 'filepath.pkl')
def test_model_datatype_error():
with pytest.raises(TypeError, match="model should be a sklearn base estimator"):
store_model(None, 'filepath')