Fuzzy Clustering Methods and their Application to Fuzzy Modeling

Uri Kroszynski, Jianjun Zhou

    Research output: Chapter in Book/Report/Conference proceedingArticle in proceedingsResearchpeer-review

    Abstract

    Fuzzy modeling techniques based upon the analysis of measured input/output data sets result in a set of rules that allow to predict system outputs from given inputs. Fuzzy clustering methods for system modeling and identification result in relatively small rule-bases, allowing fast, yet accurate prediction of outputs. This article presents an overview of some of the most popular clustering methods, namely Fuzzy Cluster-Means (FCM) and its generalizations to Fuzzy C-Lines and Elliptotypes. The algorithms for computing cluster centers and principal directions from a training data-set are described. A method to obtain an optimized number of clusters is outlined. Based upon the cluster's characteristics, a behavioural model is formulated in terms of a rule-base and an inference engine. The article reviews several variants for the model formulation. Some limitations of the methods are listed. An illustrative synthetic example is analyzed, and prediction accuracy measures are compared between the different variants
    Original languageEnglish
    Title of host publicationProceedings of the 15th. International Conference on CAD/CAM, Robotics & Factories of the Future, CARS & FOF'99
    Volume1
    Publication date1999
    Publication statusPublished - 1999
    Event15th. International Conference on CAD/CAM, Robotics and Factories of the Future, CARS&FOF'99, (ISPE) - Aguas de Lindoia, State of Sao Paulo, Brazil
    Duration: 1 Jan 1999 → …

    Conference

    Conference15th. International Conference on CAD/CAM, Robotics and Factories of the Future, CARS&FOF'99, (ISPE)
    CityAguas de Lindoia, State of Sao Paulo, Brazil
    Period01/01/1999 → …

    Cite this

    Kroszynski, U., & Zhou, J. (1999). Fuzzy Clustering Methods and their Application to Fuzzy Modeling. In Proceedings of the 15th. International Conference on CAD/CAM, Robotics & Factories of the Future, CARS & FOF'99 (Vol. 1)