Tuning an RF's Hyperparameters

Machine Learning with Tree-Based Models in Python

Elie Kawerk

Data Scientist

Random Forests Hyperparameters

CART hyperparameters
number of estimators
bootstrap
....

Tuning is expensive

Hyperparameter tuning:

computationally expensive,
sometimes leads to very slight improvement,

Weight the impact of tuning on the whole project.

Inspecting RF Hyperparameters in sklearn

# Import RandomForestRegressor 
from sklearn.ensemble import RandomForestRegressor

# Set seed for reproducibility
SEED = 1

# Instantiate a random forests regressor 'rf' 
rf = RandomForestRegressor(random_state= SEED)

# Inspect rf' s hyperparameters
rf.get_params()

{'bootstrap': True,
 'criterion': 'mse',
 'max_depth': None,
 'max_features': 'auto',
 'max_leaf_nodes': None,
 'min_impurity_decrease': 0.0,
 'min_impurity_split': None,
 'min_samples_leaf': 1,
 'min_samples_split': 2,
 'min_weight_fraction_leaf': 0.0,
 'n_estimators': 10,
 'n_jobs': -1,
 'oob_score': False,
 'random_state': 1,
 'verbose': 0,
 'warm_start': False}

# Basic imports
from sklearn.metrics import mean_squared_error as MSE
from sklearn.model_selection import GridSearchCV

# Define a grid of hyperparameter 'params_rf'
params_rf = {
              'n_estimators': [300, 400, 500],
              'max_depth': [4, 6, 8],
              'min_samples_leaf': [0.1, 0.2],
              'max_features': ['log2', 'sqrt']
             }

# Instantiate 'grid_rf'
grid_rf = GridSearchCV(estimator=rf,
                       param_grid=params_rf, 
                       cv=3,
                       scoring='neg_mean_squared_error',
                       verbose=1,
                       n_jobs=-1)

Searching for the best hyperparameters

# Fit 'grid_rf' to the training set
grid_rf.fit(X_train, y_train)

Fitting 3 folds for each of 36 candidates, totalling 108 fits
[Parallel(n_jobs=-1)]: Done  42 tasks      | elapsed:   10.0s
[Parallel(n_jobs=-1)]: Done 108 out of 108 | elapsed:   24.3s finished
RandomForestRegressor(bootstrap=True, criterion='mse', max_depth=4,
           max_features='log2', max_leaf_nodes=None,
           min_impurity_decrease=0.0, min_impurity_split=None,
           min_samples_leaf=0.1, min_samples_split=2,
           min_weight_fraction_leaf=0.0, n_estimators=400, n_jobs=1,
           oob_score=False, random_state=1, verbose=0, warm_start=False)

Extracting the best hyperparameters

# Extract the best hyperparameters from 'grid_rf'
best_hyperparams = grid_rf.best_params_

print('Best hyperparameters:\n', best_hyperparams)

Best hyperparameters:
        {'max_depth': 4,
         'max_features': 'log2', 
         'min_samples_leaf': 0.1,
         'n_estimators': 400}

Evaluating the best model performance

# Extract the best model from 'grid_rf'
best_model = grid_rf.best_estimator_
# Predict the test set labels
y_pred = best_model.predict(X_test)
# Evaluate the test set RMSE
rmse_test = MSE(y_test, y_pred)**(1/2)
# Print the test set RMSE
print('Test set RMSE of rf: {:.2f}'.format(rmse_test))

Test set RMSE of rf: 3.89

Let's practice!

Machine Learning with Tree-Based Models in Python