To build decision trees for classification models, the Analytics Library function decisiontree uses the techniques described in [Quinlan], which provide the following:
- Information gain ratio
- Pruning using the gain ratio technique
- Summary report
- Scoring function
You can use the scoring function to score or evaluate a decision tree model or to generate scoring SQL for later use.
The decisiontree function offers these additional options:
- Binning numeric variables during the tree-building process
- Table showing when each scored row is linked with a leaf node and corresponding rule set