Research On Key Technologies Of Spatial Data Warehouse

With the popular use of computerized spatial data collection tools, a huge amount of spatial data has been stored in spatial databases, geography information systems, and spatial components of many relational or other spatial information repositories.The traditional data warehosue is limited when processing the spatial data. On the one hand, the groupings and the hierarchies in spatial dimensions can be numerous and unkown at design time, therefore the traditional data cube technology is not directly applicable; on the other hand, spatial measures can hardly be managed and calculated in traditional data warehouse, so them usually be translated into non-spatial measures, in which way many spatial characters are discarded.Following the trend of the development of data warehouse, the spatial data warehouse should be built to facilitate spatial data online analysis and spatial data mining. Based on the background of the spatial data warehouse system of administration for industry and commerce in Dalian City and the Dalian City spatial data clearinghouse, focusing on the characteristics of spatial data warehouse, serveral key technologies are studied, which include the spatial data warehouse framework, spatial ETL, spatial range aggregate query and the application exploring in different application fields. On the basis of these researches, this thesis designs and implements a spatial data warehouse platform, named SEISDW. The following subjects are discussed in detail.First, to study spatial ETL. The spatial data ETL (extraction, cleansing, transformation and loading) plays an important roll in spatial data warehouse. The spatial data ETL should syncretize the spatial data and non-spatial data in the process, which distinguishes it from traditional ETL. The geocoding technique which assigning geographic coordinates to the actual address is commonly used to solve the problem. To enhance the address matching ability of geocoding, we apply the dynamic programming algorithm, which contribute much to biological sequence comparison. At the same time, the gap penalty method is improved to get better result during the address matching. The study explores the method to build spatial data warehouse in those systems, which does not store and process spatial data.Second, to study efficient indexing structure to support spatial OLAP efficiently. The range aggregate query on both non-spatial dimensions and spatial dimensions is a very important operation to support spatial OLAP. We develop the indexing scheme that can effectively performrange aggregate queries on both spatial and non-spatial dimensions. The regions of non-spatial dimensions are stored only once and indexed by aggregate cubetrees, and the regions of spatial dimensions are indexed by aR-trees. The relation of them is built by the pointer, stored in cubetrees's entries, to root of corresponding aR-tree. To optimize the operation, an indexing scheme named aCR-tree and corresponding algorithms with asymptotical performance analysis are proposed based on aggregate cubetree and aR-tree. Using both synthetic and real enterprise data, we conducted experiments to demonstrate storage overhead and range aggregate query performance of the indexing scheme. The analytical and experimental results show that the costs of range aggregate queries and storage space of aCR-tree are superior to traditional storage structures.Lastly, based on the research above, to provide a general introduction of a spatial data warehouse prototype. Its applications for administration for industry and commerce and the city spatial data clearinghouse are also discussed. The thesis introduces the background softwares including data warehouse system SEI DW and geography information system SEI GIS firstly. Then, the architecture of the spatial data warehouse system is provided with following three points: extending the spatial functions of data warehouse based on ComGIS, which include Geocoding component, spatial calculate and analysis components, map display components; distribted store model for spatial data,...