Extracting fuzzy classification rules from fuzzy clusters on the basis of separating hyperplanes
Fuzzy clustering can be applied to unsupervised and even to supervised classification problems. However the description of classes in terms of cluster prototypes and multidimensional membership functions is not always suitable for interpreting the classification system. Therefore, a number of techniques to derive fuzzy classification rules from the clusters have been proposed. These approaches usually project the clusters onto single attributes and accept a loss of information leading to a less accurate classifier. Here, the authors introduce a method that avoids projection and uses the class boundaries directly to derive the classification rules from the clusters. In this way an interpretable rule-based classifier is obtained that maintains the accuracy of the original clusters. It should be emphasized that the proposed method is restricted to fuzzy c-means clusters and cannot be extended to ellipsoidal clusters as they are computed, since in this case the cluster boundaries are no longer hyperplanes. For the entire collection see [Zbl 1015.00012].
- Supervised fuzzy clustering for the identification of fuzzy classifiers
- A methodology for construction fuzzy rule-based classification systems
- Fuzzy clustering with -hyperballs and its application to data classification
- scientific article; zbMATH DE number 1016961
- scientific article; zbMATH DE number 1495500
This page was built for publication: Extracting fuzzy classification rules from fuzzy clusters on the basis of separating hyperplanes
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4443564)
