Using Data Structure Properties in Decision Tree Classifier Design

2010
Inese Poļaka, Arkādijs Borisovs

This paper studies the techniques of performance enhancement for decision tree classifiers (DTC) that are based on data structure analysis. To improve the performance of DTC, two methods are used – class decomposition that uses the structure of class density and taxonomy based DTC design that uses interactions between attribute values. The paper shows experimental exploration of the methods, their strengths and imperfections and also outlines the directions for further research.

Keywords
attribute value taxonomy, class decomposition, classification, decision tree classifiers, classifier performance enhancement

Poļaka, I., Borisovs, A. Using Data Structure Properties in Decision Tree Classifier Design. IT and Management Science. Vol.44, 2010, pp.111-117. ISSN 1407-7493.

Publication language
English (en)