The Method Research Of Mining Association Rules In Distributed Environments

Data mining is an important area in KDD, and mining association rules in large databases is a critical aspect of data mining researches. The rapid development of Internet/Intranet makes a great progress in database applications. Since the security and cost of communication and efficiency of the applications, collecting and integrating a large amount of data from Internet/Intranet sites are not practical ways . The problem of mining association rules in distributed databases arises from this situation. This dissertation proposes the C-DMA (center-distributed mining association rules) algorithm in star structure, and a method of mining multiple layers association rules in distributed databases ,a method of mining multiple layers association rules using meta-learning and adjustable method in distributed databases, based on analyses and introduction of the basic concepts and algorithms of mining association rules and mining association rules in distributed databases. After analyzing the quantitative association rules and interestingness of association rules which are encountered often in distributed association rule mining , the dissertation proposes the methods of changing the quantitative attributions into bool attributions using FCM and Gene algorithm. The dissertation brings about conclusions below:(1) The FDM and CD are main stream algorithms for mining association rules in distributed databases. These two algorithms all work on net structure networks. However , in practical applications , considering cost in constructing the networks or management in networks, users prefer staring networks to net networks which do not meet their requirements. The dissertation proposes the C-DMA algorithm to solve this problem ,based on FDM and CD .Experimental results show that the performance of the C-DMA is available and extendable.(2) In the process of mining association rules , the quantitative attributes exit in databases . How to handle these attributes affects the mining results and the efficiency .The dissertation proposes the methods of changing the quantitative attributes into bool attributes ,so that many algorithms can be used , based on enhanced FCM and the genetic method.(3) In practical applications, multiple layers concept association rules mining are often encountered. The dissertation proposes the multiple layers concept rules miningalgorithm in distributed databases, based on designing and analyzing the algorithm of mining association rules in single database.(4) It is important to enhance the efficiency in mining association rules in distributed databases. The dissertation proposes adjustable meta-learning algorithm in mining association rules in distributed databases, based on the Sampling algorithm . (5) How to evaluating the association rules mined from large databases is very critical in applications . The dissertation proposes a method to processes the association rules mined, which combines the Klementtinen theory and similarity theory based on analyzing methods about the interestingness of association rules.