A gradient boosting model

Now we'll fit a gradient boosting (GB) model. It's been said a linear model is like a Toyota Camry, and GB is like a Black Hawk helicopter. GB has potential to outperform random forests, but doesn't always do so. This is called the no free lunch theorem, meaning we should always try lots of different models for each problem.

GB is similar to random forest models, but the difference is that trees are built successively. With each iteration, the next tree fits the residual errors from the previous tree in order to improve the fit.

For now we won't search our hyperparameters -- they've been searched for you.

This exercise is part of the course

Machine Learning for Finance in Python

View Course

Exercise instructions

Create a GradientBoostingRegressor object with the hyperparameters that have already been set for you.
Fit the gbr model to the train_features and train_targets.
Print the scores for the training and test features and targets.

Hands-on interactive exercise

Have a go at this exercise by completing this sample code.

from sklearn.ensemble import GradientBoostingRegressor

# Create GB model -- hyperparameters have already been searched for you
gbr = ____(max_features=4,
                                learning_rate=0.01,
                                n_estimators=200,
                                subsample=0.6,
                                random_state=42)
gbr.fit(____)

print(gbr.score(train_features, train_targets))
____

Edit and Run Code