Determining Best Split with Target-Driven Decision Trees
When you already know what the target assignment is and want to edit the tree so it matches this target, you can use the target variable and the best split functions. After the tree is created and you select best split,
Xpress Insight takes the values from the decision tree settings to decide when to stop searching for the optimal split. The Best Split algorithm helps you create target-driven decision trees. Its objective is to identify splits for a numeric predictor that maximizes the predictor's strength with respect to the target variable, while satisfying the settings you specified, which are used to decide when to stop searching for the optimal split. The resulting list of predictors, sorted by gain in purity, helps inform your decisions about the predictors you want to insert in your decision tree.
All data items that have the role
Possible tree variable and which are classified as categorical profile variables can be used as target.
- You can explore a best split by clicking the Explore Recommended Splits button. Alternatively you can directly apply the split suggestion to the selected node either by clicking the Split button, or right clicking on the node to display the drop-down menu.
Button Function 
Explore Recommended Splits 
Split button 
Note The best split algorithm is computed using only the selected decision node.If you chose to explore the best split, a dialog box opens with the values retrieved from the best split algorithm. You can edit the split values before using them in your tree then select either Custom or Best Split.
Best split suggestion for target, TreatmentID