Uncertain groupings: Probabilistic combination of grouping data

Abstract

Probabilistic approaches for data integration have much potential [7]. We view data integration as an iterative process where data understanding gradually increases as the data scientist continuously refines his view on how to deal with learned intricacies like data conflicts. This paper presents a probabilistic approach for integrating data on groupings. We focus on a bio-informatics use case concerning homology. A bio-informatician has a large number of homology data sources to choose from. To enable querying combined knowledge contained in these sources, they need to be integrated. We validate our approach by integrating three real-world biological databases on homology in three iterations.