A hierarchical distributed data mining architecture

Current distributed data mining (DDM) systems popularly assume distributed data sources as partitions of a virtual data table and separately mine them. In fact, when there is essential difference among data sources, the assumption will fail and DDM result quality will also be damaged. For this issue, a hierarchical DDM architecture is proposed by grouping data sources according to their similarity. Ontology technology is adopted to depict the essential content of data sources and measure their similarity.