Uncovering Patterns in Mixed Spatial and Non-Spatial Datasets, Using Qualified Association Rule Mining
No Thumbnail Available
In this thesis I propose an integrative solution combining the advantages of geographical information systems (GISs) and spatial data warehouses (sDWHs). In my endeavors to circumvent the disadvantages of both technologies, I explore the possibilities to shape the integrative solution based on the architecture of the brain and consider lessons learnt from neuroscience. I arrive to the conclusion that neuroscience is helpful in defining the architectural boundaries but does not provide sufficient insights regarding the inner workings of the constituent components of the architecture. For the latter, I resort to techniques derived from database technology, Qualitative Spatial Reasoning (QSR) and data mining. I propose an augmented table structure for a spatial data warehouse (sDWH) to contain out-of-context information from GIS databases and enhance existing mining algorithms with the capability of handling original data contained in the sDWH and newly added out-of-context information. The mining algorithms are based on data mining concepts including the a-priori principle and qualified itemset generation. I find that the number of itemsets and the overall number of records in the database impact the computational complexity.
Faculteit der Sociale Wetenschappen