Studies On New Fuzzy Clustering Algorithms

Fuzzy clustering analysis is an important branch of fuzzy pattern recognition. It is an unsupervised pattern recognition method. Clustering methods have been widely applied in various areas such asa taxonomy, geology, business, pattern recongnition and image processing etc. Studing fuzzy clustering and its applications are very important. The objective of clustering is to find the data structure and also partition the data set into groups with similar individuals. Fuzzy clustering describes each sample's uncertainty and such uncertainty sometimes can reflect the real world better then other methods.Today fuzzy clustering has been developed a very big system. In practice, fuzzy relation, fuzzy similarity and goal-function clustering algorithms are most useful. Fuzzy c-means clustering algorithm is one of the earliest goal-function clustering algorithms, which has achieved much attention. However, it remains several weaknesses and insufficiency. The signification of subjection value is one of the weaknesses. The main weaknesses of fuzzy clustering and its expandedness are sensitive to initialization, slowly convergent and sensitive to noise. Presently, according to the weakness of traditional fuzzy c-means algorithm there are many kinds of new algorithms have been proposed. First of all, the traditional fuzzy c-means clustering algorithm is studied and the meaning of the membership degrees in fizzy c-means clustering is discussed again. Some c-means clustering algorithms are proposed for interval-values data sets, weighted fuzzy c-means clustering algorithm is proposed to overcome the shortcomings of fiizzy c-means clustering algorithm. Second, presents fuzzy clustering algorithm for mixed features of symbolic and fuzzy data. We give a modified dissimilarity measure for symbolic and fuzzy data and then give FCM clustering algorithms for these mixed data types. And the proposed clustering algorithm is applied to real data with feature variables of symbolic and fuzzy data. Numerical examples illustrate that the modified dissimilarity gives better results. Finally, the clustering validity problems are studied. Three type functions are proposed which based on the fuzzy partitions. Those three validity functions are proposed based on the concepts of possibilistic distribution, Shannon entropy and subsethood measure.