The DecisionTree function creates a single classification decision tree. The function supports numeric variables and categorical attributes, handles missing values during the prediction phase, and supports GINI, entropy, and chi-square impurity measurements.
You can use the Decision Tree functions to create predictions input for the Receiver Operating Characteristic (ROC) function.
The implementation of probability estimation trees (frequency based probability calculation for class assignment and laplace correction) is according to "Tree Induction for Probability based Ranking, Provost and Domingos", Provost and Domingos (2002) (http://homes.cs.washington.edu/~pedrod/papers/mlj03a.pdf).
Function | Description |
---|---|
DecisionTree | Creates single decision tree in distributed fashion, either weighted or unweighted. Outputs model table. |
Single_Tree_Predict | Applies tree model to data input, outputting predicted labels for each data point. |