cqu_2717+ATTACHMENT01+ATTACHMENT01.4.pdf (229.89 kB)
Download file

Decisions fusion strategy : towards hybrid cluster ensemble

Download (229.89 kB)
conference contribution
posted on 06.12.2017, 00:00 by Syed HassanSyed Hassan, Brijesh VermaBrijesh Verma
Clustering ensembles have renowned as a powerful methodfor improving both the performance and constancy ofunsupervised classification solutions. However, finding aconsensus clustering from multiple algorithms is a difficultproblem that can be approached from combinatorial orstatistical perspectives. We offer a new clustering strategywhich is formulated to cluster extracted mammographyfeatures into soft clusters using unsupervised learningstrategies and ‘fuse’ the decisions using majority voting andparallel fusion in conjunction with a neural classifier. Theidea is to observe associations in the features and fuse thedecisions (made by learning algorithms) to find the strongclusters which can make impact on overall classificationaccuracy. Two novel techniques are proposed for fusion,majority-voting based data fusion, and neural-based fusion.The proposed approaches are tested and evaluated on thebenchmark database— digital database for screeningmammograms (DDSM). This study compares the performanceof the proposed ensemble approach with other fusionapproaches for clustering ensembles. Experimental resultsdemonstrate the effectiveness of the proposed method onbenchmark dataset.

Funding

Category 1 - Australian Competitive Grants (this includes ARC, NHMRC)

History

Start Page

377

End Page

382

Number of Pages

6

Start Date

01/01/2007

ISBN-10

1424415020

Location

Melbourne, Australia

Publisher

IEEE

Place of Publication

USA

Peer Reviewed

Yes

Open Access

No

External Author Affiliations

Faculty of Business and Informatics;

Era Eligible

Yes

Name of Conference

Intelligent Sensors, Sensor Networks & Information Processing Conference