Induction of decision trees in a domain with continuous attributes

Abstract

Quinlan's ID3 is a popular and efficient algorithm for inducing decision trees from concept examples, where the examples are presented as vectors of attribute-value pairs. For attributes with large, continuous domains ID3 tends to generate very com­plex decision trees. This is due to: (1) an attribute selection heuristic biased towards attributes with large domains (2) strong constraints (bias) imposed on decision trees generated and (3) the fact that ID3 does not distinguish between continuous and non- continuous attributes. ID3-IV and GID3 address the first and second problem respectively. We propose CID3, a generalization of GID3, which addresses the third problem. These algorithms are compared with respect to five criteria for decision tree quality and computational efficiency. The test domain consists of normal and ab­normal electrocardiograms (ECGs) described mainly by continuous attributes. CID3, which implements the weakest bias and uses the most domain knowledge, generates a superior quality decision tree for the ECGs.