Space Data Mining And Visualization System, A Number Of Key Technology Research

Data mining is becoming an effective means to solve the problem of data rich and information poor, it is signality for effectively dealing with magnanimous geographical data and enhancing the automatic and intelligent level of data analysis while the concept, model and methods of data mining and knowledge discovering were introduced into geography analysis field.Visualization technology can provide intuitionistic data input, result output and interactive exploration ability for data mining, provide visual mining methods with apperception, perception and estimation which overcome the shortage that objects visualization has been greatly cared for but structured information description has been slightly cared for in GIS, enhance the efficiency of data mining and confidence of mining result, the integration of visualization and spatial data mining is inevitable in the research field of geoscience.In this dissertation, spatial data integration technique based on data warehouse is systemically discussed, spatial association rules, rough set and clustering algorithms are improved, some visualization methods which are suitable for above algorithms are studied, an open 'play and plug' data mining system is designed, furthermore, a series of theoretics, methods and prototype system are pointed out on base of above research.The main content and contribution of the dissertation are as follows:(1)The related concepts and theory of spatial data integration and integration model are described, integration model of multi-source spatial data is discussed. The integrative processing technique of multi-source spatial data and multi-scale is also discussed, a data integration framework based on data warehouse is proposed, a spatial OLAP tool on web is designed.(2) An effective algorithms named MBAR which is used to discovery association rules in frequent large itemsets on base of mapping mechanism, by combination with concept tree, a multiple-level spatial association rule algorithms is put forward.(3) An extension model based on dominance relation which is used in multi-criteria decision-making, reduction of attributes and core computing algorithms in the model is analyzed, a new definition named dominance discernibility matrix is proposed, corresponding core computing and of attributes reduction algorithms are point out, a method abstracting dominance rule is also proposed.(4) A spatial clustering algorithm VSG-CLUST which is based on spatial...