An engineering blog for engineers working in Audiovisuals, Telecommunications, Electronics, ICT, Computer Engineering, Multimedia and Telematics.

20 October 2011 | Posted by Redacción Ingeniería

Journal publication in "Fuzzy Sets & Systems"

CTMedia researchers Xavier Sevillano, Francesc Alías and Joan Claudi Socoró have published a paper in the Elsevier Journal on "Fuzzy Sets & Systems" under the title: "Positional and confidence voting-based consensus functions for fuzzy cluster ensembles" Abstract: Consensus clustering, i.e. the task of combining the outcomes of several  clustering systems into a single partition, has lately attracted the  attention of researchers in the unsupervised classification field, as it  allows the creation of clustering committees that can be applied with  multiple interesting purposes, such as knowledge reuse or distributed  clustering. However, little attention has been paid to the development  of algorithms, known as consensus functions, especially designed for  consolidating the outcomes of multiple fuzzy (or soft) clustering  systems into a single fuzzy partition—despite the fact that fuzzy  clustering is far more informative than its crisp counterpart, as it  provides information regarding the degree of association between objects  and clusters that can be helpful for deriving richer descriptive data  models. For this reason, this paper presents a set of fuzzy consensus  functions capable of creating soft consensus partitions by fusing a  collection of fuzzy clusterings. Our proposals base clustering  combination on a cluster disambiguation process followed by the  application of positional and confidence voting techniques. The modular  design of these algorithms makes it possible to sequence their  constituting steps in different manners, which allows to derive versions  of the proposed consensus functions optimized from a computational  standpoint. The proposed consensus functions have been evaluated in  terms of the quality of the consensus partitions they deliver and in  terms of their running time on multiple benchmark data sets. A  comparison against several representative state-of-the-art consensus  functions reveals that our proposals constitute an appealing alternative  for conducting fuzzy consensus clustering, as they are capable of  yielding high quality consensus partitions at a low computational cost. On-line publication here.

Share

Add new comment

CAPTCHA
This question is for testing whether or not you are a human visitor and to prevent automated spam submissions.
4 + 0 =
Solve this simple math problem and enter the result. E.g. for 1+3, enter 4.