核k-凝聚聚类算法

Kernel k-aggregate clustering algorithm

为解决k-means聚类算法和k-凝聚聚类算法对于非凸形状数据聚类正确率低和模糊核聚类算法(FKCM)收敛速度慢的问题,将k-凝聚聚类算法与核函数方法相结合,在高维特征空间构造了新的核聚类算法——核k-凝聚聚类算法,实现了k-凝聚聚类算法的核化.通过Matlab编程进行数值实验,证明了核k-凝聚聚类算法在聚类的准确性、稳定性、健壮性等方面较之k-means聚类算法、k-凝聚聚类算法和FKCM有一定程度的改进.

For solving the problems that the k-means clustering algorithm and k-aggregate clustering algorithm cannot correctly cluster the non-spherical shape data, and that the convergence speed of fuzzy kernel clustering method （FKCM） is lower, a new kernel aggregate clustering algorithm -- kernel k-aggregate clustering algorithm in high dimensional feature space is introduced, which combines k-aggregate clustering algorithm with kernel function method. The data experiment using Matlab shows that the kernel k-aggregate clustering algorithm has obvious improvement in accuracy, stability and robustness of clustering compared with the k-means clustering algorithm, k-aggregate clustering algorithm and FKCM.