Research On Methods And Application For Urban Spatial Data Mining

With requirement of spatial information in city development decision as a goal, under the guidance of geographical information science and urban geography and based on the present study of spatial data mining (SDM), the author deeply studies the theory, method and appliance of urban spatial data mining (USMD). The contents that studied include the systemic structure of USMD, the models of basic spatial computation, urban spatial distribution DM, urban spatial dynamic DM, urban spatial association rule mining and urban cluster spatial data mining etc. The author brings forward collectivity frame structure and some new mining methods, improves on the some present methods of SDM, carries out a large of practical experiments, preliminarily builds up a experimental system of USDM.On the aspect of basic theory: Based on the present research, the author puts forward a frame system of USDM and the entity information model of combination of location and attribute, gives the measurement of spatial distance as a basic rule of spatial computation. By extending the method of spatial weighted matrix, brings forth the conception of spatial entity association matrix and spatial entity state association matrix, gives the method of their establishment and offers new basic tools for SDM.On the aspect of urban spatial distribution DM: Adopting the conception model of combination of location and attribute, the author brings spatial coordinates, spatial relationship and attribute character into the unitive model of spatial computation, studies city's land suitableness evaluation and the method of spatial clustering in division of urban function districts and brings forward the arithmetic of classification layer. Aiming at land spatial optimization allocation, the author puts forward SGA. In this arithmetic selection, crossover and mutation operator are progressed on the space. The author defines normative relational index based on entropy and gives the method of computation, for relational measurement among multi-elements discrete spatial fields.On the aspect of urban spatial dynamic DM: To prediction and simulation of discrete-state attributes, the author establishes the predictive method of CA which has operation, that is to say, that automatic distills part rule of state conversion from the past spatial data and makes certain cell state in the future adopting the method of random experiment in the prediction and simulation accords with actual. To prediction of continuous state attributes, the author puts forward the SAR analytic method of combination of location and attribute. This method can be used in colony prediction of continuous valued attributes of spatial cell grid. To prediction of city spatial expansion, the author brings forward predicting method and computing model of point sources radiation diffusion based on the thought of regional diffusion.On the aspect of urban spatial association knowledge mining: To static association rule, according to rough set theory the author summarizes two distilling methods of spatial association rule based on reductive data and division of equivalence. To dynamic association rule, the author puts forward the method of spatial time serial information table. This method can be used in spatial time serial association rule mining widely. But these distilling methods is discrete from computation of spatial relationship, that is to say, the method of distilling rule is different from the method of spatial computation. Aiming at complex spatial association knowledge, the author researches the method of mining spatial entity association knowledge and spatial state association knowledge according to spatial association matrix. These methods are on the basis of spatial computation.On the aspect of urban cluster SDM: In this part, the author puts forward five methods. First, the author brings forward mining thoughts and the method of estimating parameter of urban cluster distribution axis which contains coordinate and attribute index. This method includes straight line, parabola and common qua...