The Application Of Rough Set And Cloud Theory In Spatial Data Mining

In the wake of developments in modern science and technology, the size and quantity of data bases are increasingly enhanced, causing the rise in the importance of spatial data mining and knowledge discovery. Yet, in the process of data mining, the existence of the massive quantity of redundant data affects our decision-making. The Rough Set Theory in terms of obtaining decision rule and classification is the most advantageous basis. Not only can it greatly reduce the original size of the data without affecting the data expression information (data reduction), but it can also generate decision rules, thereby excavating the effective patterns within the data. In addition, the Rough Collection Theory is different from other processing uncertainty question theories, such as probability math fuzzy theory, etc. it does not need to be provided with any apriori information other than the data set to be process. Nevertheless, RST requires that the values from the processed decision table be expressed in discrete data. Therefore, the data must be discreted before using RST for data mining.This thesis begins by introducing data miming, the significance of spatial data mining, its main method of use, the process of obtaining knowledge and the type of knowledge acquired. Subsequently, the concept of RST is presented, for shadowing the in-depth research to follow.Next, the paper discusses of the heart of the problem of Rough set theory in the process of data mining—attribute reduction—and gives an analysis on the present attribute reduction algorithm based on Rough set theory, as well as a comparison of the functions of all forms of algorithms. On this basis, a type of FAE algorithm is proposed and to be used before attribute reduction to make selections on optimized attributes. A classifier model—FAERS model- is also established upon this foundation based on the Rough set analysis, and its classification results has been proven effective through experiments. However, in view of the circumstances in which the Rough set is mining the simplest rule of a common decision table or all rules are NP-hard problems, this paper introduces ant colony algorithm, providing a new attribute reduction algorithm—ACR algorithm, and utilizes ant colony algorithm to guide the direction of the search in finding attribute reduction. Through concrete examples, this algorithm has been proven valid.Thereafter, this paper presents the knowledge relevant to the cloud theory and the studies on a discrete algorithm based on a type of cloud model. On the foundation of the theoretical research in the beginning, this thesis proposes a type of prototypic system of spatial data mining based on GIS and explains the operation procedures of this system through concrete examples.Finally, upon summarizing the research results and deficiencies, the prospect of spatial data mining is elaborated.