Research On SDMKD And Intelligent Spatial Decision Support Systems

The difficulty of analyzing and processing for spatial data becomes more and more complicated, because many uncertain spatial relations are born with position attributes of spatial objects, such as spatial topology relations, spatial orientation relations, spatial distance relations and the combinations among them. And these spatial relations are always hidden in spatial data by the undisclosed form. SDMKD (Spatial data mining and knowledge discovery) is emerging when abundant of spatial data and scarce of spatial knowledge coexist, which has become the hot topic of data mining area.During the development of spatial information technology, SDSS (Spatial decision support system) which integrates the advantage of data processing of GIS and model analyzing of traditional decision support system provides effective decision support for some structured spatial problems. However, SDSS is weak for deeply spatial data analyzing and support for unstructured spatial problems without the support of machine learning, knowledge discovering, and expert system etc. ISDSS (Intelligent spatial decision support system) can make up for it.This dissertation begins with solving the problem of knowledge bottleneck of IDSS, studies the problem of acquiring, expressing and reasoning of spatial knowledge, then explores the frontier of ISDSS, such as theory basis, technology supporting, reasoning mechanism and system integration.The main researching items are as followings:1. Research of K value optimization of spatial clustering. The exiting algorithm of spatial clustering can be used in the condition that k value is given. In fact, the value of k can not be given in advance. Thus the value of k needs to be optimized Based on the research of cluster validity functions given by other scholars, by building the distance cost function as the cluster validity function, this dissertation proposes the optimization study on k value of K-means algorithm in the condition that the distance cost is smallest. The algorithm decreases the searching scope of the best value of k and makes a contribution for finding the best value of spatial clustering.2. Visualization of spatial data mining and knowledge discovery. Based on the fact of dependence and developing solely of spatial data mining and visualization technology, applies visualization technology to SDMKD and makes a research on visualization of spatial data and tools of visualization data mining. At the same time, analyzes the effects of visualization for spatial data mining at different stages and appropriates of visualization for different type knowledge, and makes a research on the visualization of SDMKD based on the GIS of spatial statistical analysis andVoronoi picture.