Font Size: a A A

The Research And Implementation Of Distributed Clustering Algorithm On Adaptive Technique And Service Message Bus

Posted on:2013-11-23Degree:MasterType:Thesis
Country:ChinaCandidate:J Q LinFull Text:PDF
GTID:2248330392954376Subject:Computer software and theory
Abstract/Summary:PDF Full Text Request
With the rapid development of information and storage technologies, more and morelarge-scale distributed database systems appear in the world. Distributed data miningprovides a facility to obtain some potential and worth information from those big amountsof datasets. Distributed cluster analysis is the most important research area in distributeddata mining and one of the hottest subjects in cluster analysis field nowadays.Distributed cluster and its applications are studied in this thesis. Message bus and improvedclustering algorithm are applied to realize the distributed cluster analysis for the distributeddatasets. The main work of this thesis is showed in the following:Firstly, the backgrounds, relates work, research purpose and significance of traditionalclustering algorithm, high dimensional clustering algorithm and distributed clusteringalgorithm are introduced. And the conception, steps and measurements of the basictechnologies are presented.Secondly, after deeply investigated and analyzed the CLIQUE algorithm that is integrateddensity-based and grid-based method, DPA-CLIQUE algorithm are proposed byself-adaptive methods and distributed parallelization. For making full use of thecharacteristic of the data being processed in parallel, DPA-CLIQUE algorithm can reducethe number of dense unit and candidate dense unit largely, decrease the complexity greatly,upgrad the accuracy of cluster effectively.Thirdly, realizing the distributed clustering algorithm model, this thesis designs a platformusing the SOA and message bus technologies. This platform includes slave node, masternode and services message bus. The distributed clustering algorithm is executed in theplatform using task parallel and data parallel.And then, the thesis realizes the prototype platform by the C#technology and the interfaceof the DPA-CLIQUE algorithm by the Weka technology, and the process of distributedclustering analysis are finished.At last, according to the background of the educational system, this thesis also designs adistributed platform based on DPA-CLIQUE algorithm. It is feasible and effective tomanage the hierarchy data in educational system.
Keywords/Search Tags:Data mining, Clustering analysis, distributed clustering, SOA, Message bus
PDF Full Text Request
Related items