Research Of RS&GA Based Spatial Data Mining Technology

Spatial data mining is a branch of data Mining, and is widely used in remote sense and geographic information system. Knowledge discovered by spatial data mining in GIS can be used for comprehension of spatial data, construction of spatial repository, reorganization of spatial database, optimization of spatial query and so on. The SDSS(Spatial Decision Support System) ,which combined with GIS ,spatial data mining and other intelligence tech, play a very important role in investigating, evaluation, and monitoring of resources, city management and planning, administration, spatial decision, evaluation and predication of disaster, cadastre management and land exploitation, traffic, agriculture, police and etc. The research work of the Dissertation is involved in spatial data mining and SDSS and the main contents are as follows:Firstly,we investigate the development and the system framework of DDS (Decision Support System ), suggest a system framework for the synthetic, distributed SDSS(Spatial Data Mining System) , and then especially study the system frameworks of distributed DSS. Secondly, the basic theories and methods of rough set and genetic algorithm are introduced. Otherwise the method of getting spatial data and the spatial data pre-treating approach based on rough set are discussed.Thirdly, a improved DM algorithm,genetic-enhanced core attributes based reduct method, and its distributed algorithm are introduced. In addition, two algorithms including Graph-based candidate frequent patterns generating algorithm and Hib&Dim-FP are proposed. All the algorithms can apply to spatial data mining under the spatial data-mining model given in the article.Then, techniques of visualization about evaluating and interpreting spatial rules are studied. Combined with multi-dimension techniques of visualization and GIS, a method of evaluating and interpreting spatial rules are proposed.Lastly, a prototype system named little three gorges prototype system validates theories and methods discussed in the article. The prototype system implements three business models including substance extracting model, pseudocolor model and spatial predicting model。 Above all, a RS&GA based spatial data mining method is proposed and the feasibility of the method are validated by a prototype system. The research of the article supplies key techniques support for real implementation of the project.