Decision trees offer an array of hyperparameters to control and tune the training result. Cross-validation is the most important tool to obtain an unbiased estimate of the generalization error, which in turn permits an informed choice among the various configuration options. sklearn offers several tools to facilitate the process of cross-validating numerous parameter settings, namely the GridSearchCV convenience class that we will illustrate in the next section. Learning curves also allow for diagnostics that evaluate potential benefits of collecting additional data to reduce the generalization error.