Research >> Projects

Knowledge Discovery Approach for Interoperability among Multiple Heterogeneous Data Sources

Back to List

Advances in database and communication technologies has motivated many organizations to the creation of a more distributed and heterogeneous data environment, which consists of legacy systems, operational databases, departmental data marts, web sites and other data sources. However, an integrated data source is essential for decision support, OLAP, knowledge discovery and data mining, web information discovery and other applications that require simultaneous and transparent access to data from multiple heterogeneous data sources. This integrated data source may take the form of a conceptual view (e.g., in a federated database system) or a consolidated central data repository (e.g., in a data warehouse). A critical and perhaps the most time-consuming step in generating this integrated data source is to identify related data objects (i.e., attributes, entities and relationships) from the underlying heterogeneous data sources.

The objective of this research is to automate the identification of semantically related data objects from multiple heterogeneous data sources and therefore to reduce the amount of human interaction. We adopt a data mining approach and apply the Self-Organizing Map (SOM) technique to cluster data objects according to their semantic similarity. We combine all available information, such as data object names, s chematic information, data contents, documentation, usage data and others (e.g., business rules), to arrive at more accurate clustering results. This research is applicable in heterogeneous database integration, data warehousing and web information integration where semantic interoperability is required.

Please click HERE for a PowerPoint presentation of the project.

Please click HERE for a PowerPoint presentation of the project.

For more information, please contact us.