Experiments with Domain Knowledge in Unsupervised Learning: U sing and Revising Theories

Abstract

Abstract. Using domain knowledge in unsupervised learning has shown to be a useful strategy when the set of examples of a given domain has not an evident structure or presents some level of noise. This background knowledge can be expressed as a set of class~fication rules and introduced as a semantic bias during the learning process. In this work we present some experiments on the use of partial domain knowledge with the tool LINNEO+, a conceptual clustering algorithm. The domain knowledge (or domain theory) is used to select a set of examples that will be used to start the learning process, this knowledge has neither to be complete nor consistent. This bias will increase the quality oi the final groups and reduce the effect oi the arder oi the examples. Some measures oi stability 01 class~fication are used. The improvement oi the concepts can be used to enhance and correct the domain knowledge. A set oi heuristics to revise the original domain theory has been experimented, yielding to some interesting results.