Enrichment of Association Rules through Exploitation of Ontology Properties - Healthcare Case Study

Abstract Association rule mining as descriptive data mining category aims to find interesting patterns on data. The quality of the patterns is measured with two metrics: confidence and support. Especially in fields dealing with sensitive data, such as healthcare, the resulting patterns should be novel and interesting. To achieve that, not only the quality of the data itself should be superior, but also other additional attributes added, do support the results. That should be achieved by using Semantic Web technologies and thus enriching data used with semantic relations between properties. A hypothesis suggests that especially tackling property relations, chain property being part of the current version of the W3C Web Ontology Language (OWL), will yield better rules. To validate the hypothesis, experiments were performed on raw data, then on an older version of OWL, which does not support the chain properties and finally on the current version of language involving chain properties. Results obtained suggest that the latter produces novel rules with strong confidence and support, not encountered in former two experiments