A fuzzy comprehensive clustering method

Shuliang Wang*, Xinzhou Wang

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

5 Citations (Scopus)

Abstract

Fuzzy comprehensive evaluation cannot reasonably differentiate the close membership values, e.g. 0.70 and 0.69. When the results have to be decided on the basis of maximum fuzzy membership value, some related information among similar objects may be neglected. At the same time, supervised fuzzy clustering analysis selects the threshold according to subjective experience. But different users may give different thresholds, and different thresholds may further get different clustering results. Integrating both fuzzy comprehensive evaluation and fuzzy clustering analysis in a unified way, this paper proposes a fuzzy comprehensive clustering method based on the maximum remainder algorithms and maximum characteristics algorithms. First, the principle of fuzzy comprehensive clustering is given. Based on the membership matrix of fuzzy comprehensive evaluation, fuzzy similar matrix is generated. Then a fuzzy equivalent matrix is produced from the fuzzy similar matrix. According to the fuzzy equivalent matrix, fuzzy clustering is implemented via the maximum remainder algorithms on the basis of fuzzy confidence level. And the grades of the resulting clusters are computed by using the maximum characteristics algorithms. Finally, a case study is given on land grading in Nanning city, the results of which show the proposed fuzzy comprehensive clustering method is able to overcome the disadvantages of either fuzzy comprehensive evaluation or fuzzy clustering analysis.

Original languageEnglish
Title of host publicationAdvanced Data Mining and Applications - Third International Conference, ADMA 2007, Proceedings
PublisherSpringer Verlag
Pages488-499
Number of pages12
ISBN (Print)9783540738701
DOIs
Publication statusPublished - 2007
Externally publishedYes
Event3rd International Conference on Advanced Data Mining and Applications, ADMA 2007 - Harbin, China
Duration: 6 Aug 20078 Aug 2007

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume4632 LNAI
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference3rd International Conference on Advanced Data Mining and Applications, ADMA 2007
Country/TerritoryChina
CityHarbin
Period6/08/078/08/07

Fingerprint

Dive into the research topics of 'A fuzzy comprehensive clustering method'. Together they form a unique fingerprint.

Cite this