Modeling Linkable Multidimensional Cubes with Logical Patterns

Student thesis: Master thesis (including HD thesis)

  • Kim Ahlstrøm Jakobsen
  • Alex Bondo Andersen
4. term, Software, Master (Master Programme)
Traditional Decision Support Systems (DSS) rely on local data to answer business questions.
This limits the decision maker to a simplied world. Ad-hoc data integration enables the decision maker to answer more business questions because the local data is enriched with situationally appropriate data.
Linked data in Resource Description Framework (RDF) format is attractive as situational data because of its growing amount and its interconnectivity between sources.
Several systems have been proposed to facilitate ad-hoc data integration and there is a high potential for further research.
We explore the potential of storing the local data in RDF as well as the situational data. We show how business questions can be answered on RDF formatted cubes annotated with the QB4OLAP vocabulary, called Linkable Multidimensional (LMD) cubes.
We introduce three logical patterns for LMD cubes, which have dierent characteristics with regards to load time, storage size, and query evaluation time.
Our novel algorithm Semantic Web OLAP Denormalizer (SWOD) is used to convert cubes of one pattern to another. The well-known TPC BenchmarkTMH (TPC-H) dataset is converted to RDF and described as LMD cubes.
We evaluate our patterns on these cubes, with queries based on the TPC-H business questions.
Publication date4 Jun 2014
Number of pages81
ID: 198543100