空间数据挖掘技术及在城镇土地定级中的应用

【作者】 童文俊；

【导师】 张晶；

【作者基本信息】 首都师范大学 ， 地图学与地理信息系统， 2008， 硕士

【摘要】 现代空间数据获取技术和计算机网络等技术的迅速发展,使得地理信息系统中的空间数据急剧膨胀。虽然这些空间数据满足了人类研究地球资源和环境的潜在需求,拓宽了可供利用的信息源,但是空间数据的复杂性和大小远非一般的事务型数据所能企及,而目前的空间数据处理手段又相对落后,使得蕴含在空间数据资源中的丰富知识被迫束之高阁,为了满足人们对空间信息日益增长的高层次需求,空间数据挖掘相关理论和技术便应运而生了。空间数据挖掘是数据挖掘的一个研究分支,而空间聚类分析是空间数据挖掘的一个重要的研究领域。因此,空间聚类算法的研究一直是空间数据挖掘研究领域中一个非常活跃的研究课题,并且已经被广泛地研究了许多年,但研究的范围主要集中在基于距离的聚类分析。本文系统地研究了划分和层次等空间数据聚类的传统算法后,总结出这些聚类算法容易陷入局部最优,在实现的过程中没有考虑到保持群体对象的全局分布特性,而且对“孤立点”信息比较敏感。正是由于这些不足,大大限制了聚类算法在GIS领域当中的应用。同时遗传算法模仿生物进化过程的自然选择和进化机制,是一种基于群体的全局随机优化算法。因此,可以考虑运用遗传算法来解决空间聚类问题。本文正是将“局部和整体”两个层次进行聚类结果优化作为出发点,分析了遗传算法与常规的聚类方法各自的优点,研究了基于遗传算法的空间聚类方法。经过具体的理论分析和模拟试验,发现该方法是可行的,而且在具体的试验过程了,确实达到了理论的要求。最后,针对目前土地定级划分中人为的主观因素占主导的现实状况,利用改进的空间聚类算法,以基本地价区片为样本,进行土地等级划分,取得了比较理想的结果。本文的研究还存在一些不足之处,文章对这部分提出了展望,以推动进一步的工作。更多还原

【Abstract】 Modern spatial data acquisition technology and computer network technology is developing rapidly,making GIS spatial data in the rapid expansion.Although these spatial data meets the human’s potential demand for study of Earth’s resources and environment,and broadens the sources of information available,the complexity and size of the data of general affairs is far from the spatial data.Currently,space data processing means is relatively straggly,so it makes spatial data contained in the wealth of knowledge resources to be shelved.In order to meet the growing demand of high-level information for spatial data,spatial data mining theory and the technology has come into being.Spatial Data Mining is a research branch of data mining,and the spatial clustering analysis is an important area of research of spatial data mining.Spatial clustering algorithm has been a very active research topic on spatial Data Mining Research field,and has been widely studied for many years. However,the scope of the study mainly concentrated in the cluster analysis based on distance.In this paper,a systematic study of traditional clustering algorithms of spatial data is made,such as hierarchical clustering algorithm,and we conclude that this clustering algorithm is summed up easily into local optimum.In the process of realization does not take into account the overall situation of target groups to maintain distribution,and is more sensitive to the isolated information.Because of these deficiencies,it greatly limits the clustering algorithm in the field of GIS Application.Meanwhile,genetic algorithms mimic biological evolution process of natural selection and evolution mechanism,and it is a global optimization algorithm based on a random group.So we can be considered the use of genetic algorithms to solve the problem of spatial clustering.Considering the relationships between party and entirety of Clustering object’s feature,we came out of a method of combining the genetic algorithm and the conventional clustering,to design a spatial clustering algorithm based on genetic algorithm.After the specific theoretical analysis and simulation test,found that the method is feasible,and indeed reached a theoretical requirements in the course of specific test.Finally,we analysis the status of distinction of grading of the land,and concludes that artificial subjective factors is a dominated factor.using improved of spatial clustering algorithm,as the basic premium for the sample area,we do the work of Land Classification,and get the fairly satisfactory results.Although there are some fields not to be studied,this will be the important research area for the future.更多还原