Study On Techniques Of Spatial Database Based-data Mining

In this paper, data mining techniques based on spatial database,especially GIS database, are studied, of which spatial data classification based on rough set theory and spatial association rule mining are studied in detail. The differences between mining in spatial database and in relational database are analyzed. Rough set theory is applied in spatial data classification, and during classification spatial and non-spatial data are processed by rough set theory simultaneously.In spatial classification, the spatial attributes and non-spatial attributes are considered at the same time. To make classifying with rough set theory possible, the spatial predications relative to classification task and attributes in original attribute table must be filled to decision table firstly, and then classification with rough set theory is performed on the decision table: attribute discretization of decision table, attribute reduction and value reduction of decision table. In this paper, a novel decision table discretization algorithm is presented, which has fine attribute reduction function in time of data discretization and increases quality of classification. In the mining of spatial association rules, a simple method defining transactions in spatial data is adapted, which can keep the location relationship among spatial data and is easy to compute.On the basis of studying spatial data mining method, a spatial data mining system is designed and realized, which can obtain spatial association rules and spatial classification rules from GIS database. The system consists of two modules engaged in mining spatial association rules and spatial classification rules separately. The spatial association rule mining module can discover the location association among geographic objects in a map. The spatial classification module can classify the geographic objects according to a pre-specified attribute, and the classification result can be used to provide decision support. The classification result can be saved to and loaded from an XML file.