Methods And Applications Study Of Cluster-based Spatial Data Mining

With the fast development of spatial data obtained technology, spatial data increase up rapidly. There are a great many intensive knowledge and law to excavate in the spatial data databases. Cluster analysis is one of important means, which requires algorithm with high efficient and auto ascertain or user easy ascertain the demanded parameters. Therefore, our study mostly surrounds 3 aspects as follows.1.The hierarchical cluster algorithm and applicationFive space regular clustering algorithm's performance summarization was attained through many times of attempt, compare;A method of the result validity through dispersion analysis idea was introduced, it makes the result more available;A method of proceeding data in advance was introduce, it acquirs more quickly operational speed, and it is able to do discrete and continuous variable, to auto-select the quantity of clustering, to proceed magnanimity data.2.The k — means cluster algorithm and applicationA method of proceeding all variable was introduced by optimizing scale transform;A random sampling algorithm was introduced, we made use of HCA to acquire the best categorical number, considered fully data meaning and distribution, so the selected initial clustering centroid were more representative;A touched algorithm was introduced base on discussion, it was reasonable in theory and feasible in practice, and improved the stability of result, reduced the dependence to initial clustering centroid.3. Cluster-based spatial data mining systemA spatial data mining system frame was introduced, consist of 3 part: aim, design and realization. Adopt modularization design idea, system design was divided into 4 module: data access, clustering, user mutual and knowledge management;The clustering method was integrated and system supplied technology support for spatial data mining means and application.In a word, research of cluster-based spatial data mining technology can improve GIS query and parsing technique to a new phase of knowledge discovery , and erect intellectualized GIS from among discoverable knowledge, supply valuable knowledge for decision-making, bring inestimable benefit. Hence, methods and applications study of cluster-based spatial data mining are significance in theory and practice.