The Research Of Mining Association Rules And It's Application On Remote Sense Image

With the widespread computerization in business, government, and enterprise, the efficient and effective discovery of validity information form large databases becomes essential. Data Mining, also named as Knowledge Discovery in Database, is an advanced handle process, in which we can pick up many novel, validity and readable patterns form very large amounts of data.Association Rules is one of the most important branches of data mining research works, its target is to research the interested relationship in some affairs. It can bring some Association Rules, what we can use for finding different relationships among some deals, etc. In this thesis we have done some work as follows:The basic concepts and theories of Association Rules are described. The typical algorithms: Apriori and FP-growth of mining association rules are discussed and analyzed in merits and defects of these algorithms. Because the Apriori algorithm must scan the database for many times, this thesis proposes a improve algorithm based on a property of the Apriori. We can reduce the times scanning the database by estimating the unit of the k-1 dimension subset. The improve algorithm on FP-growth is brought forward and programmed to execute. When the database become very huge, we will meet some bottlenecks because of the calculating the times frequent- set inserted in the FP-tree. This thesis proposes an algorithm which can fast constructing the FP-tree based with another small database.We transform the raw transaction data with quantitative attribute into fuzzy - set transaction data. Firstly, we can get the center words by clustering, and then we can establish the fuzzy membership function, which can be used in calculating the membership. At last, we use the max one instead of the formerly one.The characteristics of remote sense image data are analyzed. Because of problems existing in the application of association rules in remote sense image and mining, an improved method including fuzzy-set, Association Rules, and clustering is introduced to solve the problem.