We will further fine tune our models using hyper parameter tuning through GridSeachCV. It loops through predefined hyperparameters and fit your estimator (model) on your training set. After this you can select the best set of parameters from the listed hyperparameters to use with your model.
First we will import GridSearchCV from Scikit-learn
from sklearn.model_selection import <<your code goes here>>
Then we will define a set of various n_estimators and max_features in your model. First it will try a set of 3 n_estimators and 4 max_features giving a total of 12 combination of parameters. Next it will set the bootstrap hyperparameter as False and try a combination of 6 different hyperparameters as shown below
param_grid = [
{'n_estimators': [3, 10, 30], 'max_features': [2, 4, 6, 8]},
{'bootstrap': [False], 'n_estimators': [3, 10], 'max_features': [2, 3, 4]},
]
Now we will use these combination of hyperparameters on our Random Forest model
forest_reg = RandomForestRegressor(random_state=42)
grid_search = GridSearchCV(forest_reg, param_grid, cv=5,
scoring='neg_mean_squared_error',
return_train_score=True)
grid_search.fit(housing_prepared, housing_labels)
Now let's see the best combination of parameters
grid_search.best_params_
And the best combination of estimator
grid_search.best_estimator_
Finally, let's computer the results and print the scores
cvres = grid_search.cv_results_
for mean_score, params in zip(cvres["mean_test_score"], cvres["params"]):
print(np.sqrt(-mean_score), params)
Loading comments...