Entropy measures of flow graphs with applications to decision trees

Abstract

Entropy is a fundamental principle in many disciplines such as information theory, thermodynamics, and more recently, artificial intelligence. In this article, a measure of entropy on Pawlak's mathematical flow graph is introduced. The predictability and quality of a flow graph can be derived directly from the entropy. An application to decision tree generation from a flow graph is examined. In particular, entropy measures on flow graphs lead to a new methodology of reasoning from data and shows rigorous relationships between flow graphs, entropy and decision trees. © 2009 Springer Berlin Heidelberg.