Angoss Decision Trees provide an interactive and intuitive interface for building and exploring segments, discovering relationships between variables, performing variable reduction and much more. They provide ultimate flexibility both on the tree and individual node levels.
Decision Trees can be grown manually by creating splits with custom settings, automatically by using segmentation algorithms or through a combination of methods.
Aside from being a self-sufficient decision support tool, Decision Trees can be used to accelerate modeling creation processes:
- Select the most predictive variables for further modeling with regression or other models
- Quickly perform coarse classing of variables with optimal cut-off points for scorecards
Using automatic code generation, Decision Trees can be converted to predictive rules that can be deployed in other environments. Supported code types include SQL, SAS, Java and PMML. Decision Trees can also be deployed directly through the KnowledgeSEEKER scoring interface.
Decision Trees employ a rich set of features:
| Decision Trees Feature Highlights |
| Discrete (categorical) and continuous dependent variables |
| 4 algorithms based on improved variants of CHAID and CART algorithms |
| Advanced algorithm control and user preferences to fine-tune tree growth, for example binary and n-ary tree options, value grouping types and open intervals |
| Simple copy trees and splits (subtrees) within the same tree or between different trees |
| Node Report that display node rules and statistics for every decision (terminal) node |
| An Interactive Tree Map that is synchronized with the main tree view |
| Dynamic node-monitoring charts to view the distribution of any variable within any node, synchronized with the main tree view |
| Easy copying of tree images and underlying data to Microsoft Office® applications |
|
Automatic generation of deployment code in several languages: structured English, generic code, SAS, SPSS, SQL, Java, PMML and XML |