Fuzzy clustering based on cooccurrence matrix and its application to data retrieval

Kohei Inoue, Kiichi Urahama

Research output: Contribution to journalArticlepeer-review

3 Citations (Scopus)

Abstract

A fuzzy clustering method is proposed to cluster objects and classes based on the cooccurrence matrix that represents the cooccurrence relationship of the objects and the classes. It is a type of method known as a graph spectral method that reduces the problem to an eigenvalue problem and successively extracts the clusters. A method based on the similarity matrix is applied to the cooccurrence matrix and is extended to hierarchical fuzzy clustering. This method obtains the cluster information of the class simultaneously with object clustering. As an application example of this clustering method, we present data retrieval by key words. Since clustering extracts the overall data structure to some degree, the retrieval is robust in noisy data similar to Latent Semantic Indexing. Fuzzy clustering performs object-level retrieval because the detailed information lost in hard clustering is preserved.

Original languageEnglish
Pages (from-to)10-19
Number of pages10
JournalElectronics and Communications in Japan, Part II: Electronics (English translation of Denshi Tsushin Gakkai Ronbunshi)
Volume84
Issue number8
DOIs
Publication statusPublished - Aug 1 2001

All Science Journal Classification (ASJC) codes

  • Physics and Astronomy(all)
  • Computer Networks and Communications
  • Electrical and Electronic Engineering

Fingerprint Dive into the research topics of 'Fuzzy clustering based on cooccurrence matrix and its application to data retrieval'. Together they form a unique fingerprint.

Cite this