Applying Category Utility and Partition Utility Functions
to Semantic-Based Conceptual Clustering Approach
Ivo Marinchev
Institute of Information Technologies, 1113 Sofia
Abstract:
The paper presents an instance of our general semantic-based approach to creating concept hierarchies [7] particularly tuned for better performance on data represented with attribute-value and atomic formulae languages. We achieve this goal by incorporation of category utility and partition utility functions in its distance measure. Finally, we compare the performance of augmented algorithm to that of the COBWEB implementation given in the WEKA 3 ML system.
Keywords: lattice theory, distances, category utility, partition utility, hierarchical clustering, least general generalization (lgg).