Practical Predictive Analytics

Other options to render decision trees

These are the important control parameters that are used by rpart to grow a tree. You can often play with these values to get a tree to render in a reasonable form, and to control its complexity:

• cp parameters: The cp (complexity parameter) is an option used in rpart that controls how liberal you want the splitting algorithm to be when it decides whether or not to split a node. Utilizing a small cp value (< 0.01) can generate an enormous tree that can contain unexplainable splits, while a tree with a high cp value (> 0.05) can produce a tree containing only obvious information. Therefore, it is important to set the cp level at a number that generates a fair amount of nodes, yet remains explainable. ...