On Improving Performance and Increasing Useability of EFS

  • Edwin Lughofer
Part of the Studies in Fuzziness and Soft Computing book series (STUDFUZZ, volume 266)


This chapter deals with further improvements of evolving fuzzy systems with respect to performance and usability. Opposed to the previous section, which demonstrated aspects which are necessary (’must haves’) for guiding evolving fuzzy systems to more robustness, process safety and meaningful output responses in case of specific occasions in the data streams (drift, outliers, faults, extrapolation situations etc.), the issues treated in this chapter are ’nice to have’ extensions for further increasing prediction power and usability. To achieve this, we demonstrate the following strategies:
  • Dynamic rule split-and-merge strategies during incremental learning for an improved representation of local partitions in the feature space (Section 5.1).

  • An on-line feature weighting concept as a kind of on-line adaptive soft feature selection in evolving fuzzy systems, which helps to reduce the curse of dimensionality dynamically and in smooth manner in case of high-dimensional problems (Section 5.2).

  • Active and semi-supervised learning the reduce labelling and feedback effort of operators at on-line classification and identification systems (Section 5.3).

  • The concept of incremental classifier fusion for boosting performance of single evolving fuzzy classifiers by exploiting the diversity in their predictions (Section 5.4).

  • An introduction to the concept of dynamic data mining, where the data batches are not temporally but spatially distributed and loaded (Section 5.5.1).

  • Lazy learning with fuzzy systems: an alternative concept to evolving fuzzy systems (Section 5.5.2).

Although these aspects can be seen as typical ’add-ons’ for the basic evolving fuzzy systems approaches demonstrated in Chapter 3, they may be key stones in some learning context, for instance when dealing with very high-dimensional problems or feedback from experts/operators with different experience or confidences in their decisions.


Fuzzy System Membership Degree Unlabelled Data Feature Weight Cluster Partition 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.


Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

Copyright information

© Springer-Verlag Berlin Heidelberg 2011

Authors and Affiliations

  • Edwin Lughofer

    There are no affiliations available

    Personalised recommendations