【作者】 周海燕；

【导师】 王家耀；

【作者基本信息】 中国人民解放军信息工程大学 ， 地图制图学与地理信息工程， 2003， 博士

【摘要】 空间数据挖掘是指从空间数据库中提取用户感兴趣的空间模式与特征、空间与非空间数据的普遍关系及其它一些隐含在空间数据中的普遍的数据特征。本文系统研究了空间数据挖掘的理论、方法和应用。主要内容有： (1)建立起空间数据挖掘的基础理论和技术框架，进一步完善了空间数据挖掘的理论和方法。阐述了空间数据挖掘的定义与特点，提出一种包括数据源、挖掘器、用户界面三层结构的空间数据挖掘体系结构，阐述空间数据挖掘的基本步骤和从空间数据库中能发现的九种知识类型，系统研究了17种空间数据挖掘方法，阐述了各种方法的特点和适用范围，阐述了空间数据挖掘与其它相关学科的区别与联系，指出空间数据挖掘的主要研究方向，提出开发空间数据挖掘系统的几条原则。 (2)将粗集理论引入空间数据挖掘领域，系统地研究了粗集理论用于空间数据挖掘的基础理论和方法，包括粗集的基本概念和性质、粗集的扩展模型、空间数据库中知识表达系统的分类、属性表的一致性分析、属性的依赖关系和属性的重要性、决策表属性约简和属性值约简等。 (3)提出空间关联规则主要指空间对象之间的空间和非空间关系，指出空间关联规则的形式十分丰富，重点研究了两种形式的空间关联规则的挖掘。阐述了A=>B[s％，c％]形式的空间关联规则的基本概念和算法，详细研究了一种逐步求精的空间关联规则挖掘算法的实现；提出一种基于空间数据立方体的空间关联规则挖掘的新思路；将空间统计分析引入空间关联规则挖掘领域，研究了空间权重矩阵、空间自相关、空间关联等的度量函数，并利用空间统计分析技术发现空间相关关系和空间关联规则。 (4)系统研究了七种典型的空间数据聚类方法，积极探索基于约束条件的空间聚类问题的解决方案；将遗传算法引入空间数据聚类领域，提出一种基于遗传算法的空间聚类算法，该算法兼顾了局部收敛和全局收敛性能。 (5)系统研究了Voronoi图和Delaunay三角网的定义、性质及各种建立算法，并对它们在空间数据挖掘中的应用进行了探索性研究：提出Voronoi图是界定空间目标(现象)的空间影响范围的一种行之有效的办法；从理论上论证了Voronoi图的形成与城市中心地的形成是一致的，提出Delaunay三角网是建立城镇网络体系的最佳模型；研究了利用Voronoi图进行公共设施选址优化的算法及实现。更多还原

【Abstract】 Spatial data mining (SDM) refers to picking up interesting rules from spatial database, such as spatial patterns and characteristics, the universal relations of spatial and non-spatial data and other universal implicated in spatial data. This thesis studies on the theories, techniques and the applications of SDM. The main content of this paper include the follows:(1) The basic theory and technology framework of spatial data mining are established and the theory and methods are perfectly developed. The definition and characteristics of SDM are set forth, and a structure of spatial data mining system including data source, miner and user interface is put forward. The essential processes of SDM are studied and nine types of rules resulting in mining are discussed. There are 17 kinds of spatial data mining approaches researched in this paper and each method’s characteristics are analyzed. Moreover, the difference and relationship between SDM and other related subjects are discussed in detail. The future research directions of SDM and some principles on developing SDM system are pointed out.(2) The theory of Rough Sets is introduced into SDM domain. The basal theory and techniques of Rough Sets including the basic notion and character, the extended models, the classification of knowledge expression system in spatial database, the coherence analysis of attribute table, the relying relations between attributes, the importance of attribute, the reduction of attribute and the reduction of attribute value in decision table are studied by the numbers.(3) The definition of spatial association rule is defined as the spatial and non-spatial relations between spatial objects. The forms of spatial association rule are abundant. Two important types of spatial association rule are studied. Firstly, The notion of the form as A=>B[s%, c%] is researched and some algorithms are discussed. An algorithm named A Progressive Refinement Approach to Spatial Data Mining is discussed in detail. And a new thought of mining spatial association rule based on spatial data cube is brought forward. Then, the spatial statistical analysis techniques are introduced into SDM domain. The measurement functions of spatial weight matrix, spatial auto-correlation and spatial association are studied.(4) Seven kinds of spatial data clustering approaches are studied. And the technique to solve the problem of Constraint-based Spatial Cluster Analysis is explored. In addition, a new spatial clustering algorithm based on Genetic Algorithms is set forward and it can give attention to local constringency and the whole constringency.(5) The definitions, characteristics and all kinds of building algorithms of the Voronoi Diagram and the Delaunay Triangle are introduced. Their applications in SDM are explored. That the Voronoi Diagram is an effective method to partition the influence regions between spatial objects and phenomena is put forward, and that the principle of building Voronoi Diagram is identical to the forming central place is proved. Then, that the Delaunay Triangle is the best model to set up the cities network is brought forward. Finally, the problem of spatial establishment location selection by means of the Voronoi Diagram is studied.更多还原