Augmenting decision tree models using self-organizing maps

Abstract

This study considers the application of the Self-Organizing Map technique on a decision tree model generated to achieve model-augmented visualization, based on a visual perception model scheme called VAM-DM. It supports the visual analysis of a data mining model in the adjustment phase, also combining complementary views of graphical artifacts for each component or node of the decision tree. It seeks to answer user generic questions regarding the model inner workings and to achieve a better understanding of the model finally obtained. In this context, the Self-Organizing Map technique serves a dual purpose: spatial partition of the data subset associated with a tree node and partition visualization with a map. Finally, a controlled experiment is carried out with a software prototype and two user groups, novices and experts in DM's processes, and results from this experiment are analyzed. This analysis allows us to assess the usefulness of the Self-Organizing Map technique for augmented decision tree model and their efficiency to support the comprehension of the generated model. © 2013 Springer International Publishing.