【作者】 邓有莲；

【导师】 周定康；

【作者基本信息】 江西师范大学 ， 计算机应用技术， 2007， 硕士

【摘要】 Tobler的第一地理规则:“所有的事物都是有联系的,一个地方发生的事件总是与它附近发生的事件有关联,并且相距近的事物之间的联系一般比相距远的事物之间的联系要紧密。”它指出了空间对象之间具有空间依赖性,由此产生了在GIS数据库中挖掘空间关联规则。本文使用了最小包围矩形、平面扫描算法来计算对象间的邻域关系。另外,在GIS数据库中,对象间的空间关系是隐含的且计算复杂度高,本文根据扩展的空间谓词概念层次树从上至下逐一计算,直到所有层次的结点被计算完毕或被剪枝。单个谓词采用空间关系判定方法。这样,空间数据转换为事务数据。对于在事务数据库中挖掘关联规则,如果根据应用领域确定一个挖掘目标,即确定关联规则的条件属性和决策属性,可以收到事半功倍的效果。因此,本文重点研究了在GIS数据库中挖掘带有决策属性集的空间关联规则的过程与方法。在挖掘关联规则过程中,本文首先将事务数据按各属性进行不可区分关系划分。在此基础上,通过区分矩阵方法对条件属性进行绝对约简,同时设计了求绝对约简的算法;而且,设计出了具有决策属性集可行的多层多维关联规则挖掘算法。更多还原

【Abstract】 Content:It is Tobler’s first geographical rule that everythings have associations, the incident is always associated with incidents which took place nearby,and more close is the distance between place,more close is the associations. This rule pointed out that spatial dependence exists in spatial objects. Therefore, mining spatial association rules in GIS database is suggested. In this paper, mbr and plane sweep algorithm are used to compute neigbour relation between spatial objects. Besides,spatial relations between spatial objects are covered , if they are computed, it will cost much time and memory. To solve this problem, spatial relations are computed gradually from up to down according to concept hierarchic tree of spatial predication , until all of node are computed or cut. Spatial relation discriminant method is used to compute single node. Thus, spatial data is transformed into transaction data.As for mining association rules in the transation database, if a mining target is decided according to application domain, that’s to say ,determine condition attributes and decision attributes ,then the work will have half effort and double results. Hence, this paper mainly studies the techniques on process and methods of mining association rules with decision attributes in GIS database.In the process of mining association rules, transaction data are divided firstly according to indiscernibility relation of attribute. On the basis of the first step, discernibility matrix is used to calculate absolute reduction and absolute reduction algorithm is designed in this paper. Mining multi-layer and multi_dimension association rules algorithm with decision attributes is also designed on the basis of equivalence of reduction attributes.更多还原