Graph-Based Data Warehousing Using the Core-Facets Model.

There are a growing number of data-mining techniques that model and analyze data in the form of graphs. Graphs can link otherwise disparate data to form a holistic view of the dataset. Unfortunately, it can be challenging to manage the resulting large graph and use it during data analysis. To facilitate managing and operating on graphs, the Core-Facets model offers a framework for graph-based data warehousing. The Core-Facets model builds a heterogeneous attributed core graph from multiple data sources and creates facet graphs for desired analyses. Facet graphs can transform the heterogeneous core graph into various purpose-specific homogeneous graphs, thereby enabling the use of traditional graph analysis techniques. The Core-Facets model also supports new opportunities for multi-view data mining. This paper discusses an implementation of the Core-Facets model, which provides a data warehousing framework for tasks ranging from data collection to graph modeling to graph preparation for analysis.