An ontology-based method to link database integration and data mining within a biomedical distributed KDD

Abstract

Over the last years, collaborative research has been continuously growing in many scientific areas such as biomedicine. However, traditional Knowledge Discovery in Databases (KDD) processes generally adopt centralized approaches that do not fully address many research needs in these distributed environments. This paper presents a method to improve traditional centralized KDD by adopting an ontology-based distributed model. Ontologies are used within this model: (i) as Virtual Schemas (VS) to solve structural heterogeneities in databases and (ii) as frameworks to guide automatic transformations when data is retrieved by users-Preprocessing Ontologies (PO). Both types of ontologies aim to facilitate data gathering and preprocessing while maintaining data source decentralization. This ontology-based approach allows to link database integration and data mining, improving final results, reusability and interoperability. The results obtained present improvements in outcome performance and new capabilities compared to traditional KDD processes. © 2009 Springer Berlin Heidelberg.