Tuning the Decision Tree Algorithm

Fine tune the Decision Tree algorithm with various parameters.

The Decision Tree algorithm is implemented with reasonable defaults for splitting and termination criteria. However several build settings are available for fine tuning.

You can specify a homogeneity metric for finding the optimal split condition for a tree. The default metric is gini. The entropy metric is also available.

Settings for controlling the growth of the tree are also available. You can specify the maximum depth of the tree, the minimum number of cases required in a child node, the minimum number of cases required in a node in order for a further split to be possible, the minimum number of cases in a child node, and the minimum number of cases required in a node in order for a further split to be possible.

Note:

The term hyperparameter is also interchangeably used for model setting.

The training data attributes are binned as part of the algorithm's data preparation. You can alter the number of bins used by the binning step. There is a trade-off between the number of bins used and the time required for the build.

See Also:

DBMS_DATA_MINING —Algorithm Settings: Decision Tree for a listing and description of the available model settings.

Note:

The term hyperparameter is also interchangeably used for model setting.