Clustering objects with degree of classification

This paper proposes a fuzzy clustering method under the intrinsically classified structure of data through dissimilarity of objects at each variable. In order to extract the classification structure, the variable-based fuzzy clustering method is exploited and the degree of classification for each object with respect to each variable is defined. This degree shows individually classified power of an object with respect to a variable. By applying this degree to the data, a stable classification solution which is not sensitive to the outlier is obtained. Several numerical examples show the improved performance and the applicability of our proposed method.