Robust Fuzzy Clustering Using Adaptive Fuzzy Meridians
The fuzzy clustering methods are useful in the data mining applications. This paper describes a new fuzzy clustering method in which each cluster prototype is calculated as a fuzzy meridian. The meridian is the maximum likelihood estimator of the location for the meridian distribution. The value of the meridian depends on the data samples and also depends on the medianity parameter. The sample meridian is extended to fuzzy sets to define a fuzzy meridian. For the estimation of medianity parameter value, the classical Parzen window method by real non–negative weights has been generalized. An example illustrating the robustness of the proposed method was given.
KeywordsFuzzy Cluster Laplace Distribution Medianity Parameter Partition Matrix Fuzzy Cluster Method
Unable to display preview. Download preview PDF.