| Pruning of trees | Focusing on important outcome groups | |
The features offered by a MinedTree object make it relatively easy to obtain tree profiles from data source tables. The important task is then to interpret the resulting trees in order to understand the patterns in data and apply these rules and patterns in future decision making. Every leaf in the tree represents a rule, pattern or profile which relates an outcome group to a set of attribute values. The outcome group assigned to the leaf rarely has a probability of one. This means that other outcome groups also have a probability of belonging to this particular profile.
When deeper understanding of the resulting tree is required beyond the simple profiling of data then we recommend that you split your analysis into a number of analyses each with two outcome groups only. For example, if your analysis contain four outcome groups G1, G2, G3 and G4 then consider splitting it into four separate analyses. Each of the individual analyses will have one of the groups against the other three. This will simplify the understanding and interpretation of the resulting trees and verification, since it allows us to focus on one outcome group at a time.