Inducing Decision Trees via Concept Lattices

The paper presents a new method of decision tree induction based on formal concept analysis (FCA). The decision tree is derived using a concept lattice, i.e. a hierarchy of clusters provided by FCA. The idea behind is to look at a concept lattice as a collection of overlapping trees. The main purpose of the paper is to explore the possibility of using FCA in the problem of decision tree induction. We present our method and provide comparisons with selected methods of decision tree induction on testing datasets.

• Type: Conference
• Year: 2007
• Where: CLA
• Authors: Radim Belohlávek, Bernard De Baets, Jan Outrata, Vilém Vychodil
• Keywords: Artificial Intelligence, CLA 2007, Concept Lattice, Decision Tree, Decision Tree Induction