Efficient knowledge discovery through the integration of heterogeneous data

TLDR: By integrating heterogeneous data it is often possible to discover new information at a finer level of granularity than that available in any of the contributing data sources, and a system may be enabled to induce new rules based on information at that would be possible without integration.

Abstract: It is commonly the case that a distributed database holds data originating from a number of different sources. These heterogeneous data sources may provide different views of the same data, or they may be different samples from the same population. In each case, a methodology is provided for combining data which may be held at different levels of granularity. Variations in granularity arise due to the use of different levels within a concept hierarchy or the use of different concept hierarchies. Data integration is accomplished using the intersection hypergraph to produce the integrated universal classification scheme and to determine the cardinalities of each category within the universal table. In the commonly occurring case of continuous or ordinal data, an explicit and efficient computational algorithm is presented. By integrating heterogeneous data it is often possible to discover new information at a finer level of granularity than that available in any of the contributing data sources. A system may be thus be enabled, so as to induce new rules based on information at a finer level of granularity than that would be possible without integration.