Search

Nonlinear clustering has attracted an increasing amount of attention recently. In this paper, we propose a new nonlinear clustering method based on crowd movement and selection (CCMS). Different from most of the existing clustering methods that concentrate on cluster centers, our method focuses on the data points themselves and the data distribution of their neighborhood. Based on this novel idea, some useful rules are designed to transform the original nonlinearly separable datasets into linearly separable datasets. Finally, a new clustering phase is designed to partition the transformed data. Extensive experiments have been conducted to evaluate the effectiveness of the proposed CCMS method. [ABSTRACT FROM AUTHOR]