A Robust Fuzzy Kernel Clustering Algorithm

Traditional fuzzy kernel clustering methods does Iterative clustering in the original data space or in the feature space by mapping the samples into high-dimensional feature space through a kernel function These methods with normalized fuzzy degree of membership has weak robustness against noises and outliers, and lack of effective kernel parameter selection method. To overcome these problems, a robust kernel clustering algorithm is proposed to enhance the robustness by using typical parameter. Meawhile, a kernel function parameter optimization method under the unsupervised condition is also proposed in this paper. The experimental results show that the new algorithm is not only effective to the linear inseparable datasets with noisy data, but also more robust compared with other similar clustering algorithms and can obtain better clustering accuracy under noise jamming.