Many Faces of a Tree
Simon Urbanek · OPUS (Augsburg University) · 2003
Tree based models represent an appealing alternative to other commonly used methods for classification and regression tasks. Beside their statistical properties one of the most valued features of tree models is the straight-forward interpretability. Treemodels can be often easily communicated to the data set owner who may have less profound statistical knowledge. In order to properly analyze and interpret a tree model, it is important to look at the model from different views and consider the underlying data. In this paper we will present several ways of looking at a tree model, some of them less known. Unfortunately most software tools used for tree model building lack the features necessary for visualization and deeper analysis of the models. In this paper we will also briefly illustrate how our research software, KLIMT (Klassification Interactive Methods for Trees), can be used for interactive graphical analysis of trees using these extended visualization methods.