SOIM: Similarity Measures on Ontology Instances Based on Mixed Features

摘要

Clustering has been widely applied to various domains to explore the useful patterns inside data. Clustering quality can be improved using Knowledge represented by ontology. Nevertheless, most traditional ontology-based clustering algorithms are limited to handle categorical instances. But in real case study, ontology contains both numerical and categorical attributes. In this paper, we propose a new method for clustering knowledge contained in the ontology based on mixed features. The main contribution is the proposition of new similarity measures that combine numerical and nominal variables along different dimensions (instances, attributes, and relation-ships). Three kinds of similarity measures are so defined: instances-based similarity IS, relations-based similarity RS and attributes-based similarity AS. These three measures are then combined into an overall similarity measure. This combined measure is used for clustering.