基于核的直觉模糊聚类算法

Kernel based intuitionistic fuzzy clustering algorithm

针对现有的直觉模糊聚类算法性能的问题,提出一种基于核的直觉模糊聚类算法(IFKCM)。该算法引入高斯核函数,将直觉模糊集合从原始观察空间映射到高维特征空间,减少了计算时间且提高了聚类精度;同时改进了现有的直觉模糊聚类算法中的概率型约束条件,使其对噪声和野值点具有较好的鲁棒性。最后,通过实际数据和人工数据与常用聚类算法进行了对比实验,结果表明该算法较大幅度地提高了直觉模糊聚类算法的性能。

A kernel based intuitionistic fuzzy clustering algorithm named IFKCM was proposed on the basis of analyzing the deficiency of the existing clustering algorithm. The new algorithm, through introducing Gauss kernel, mapped the intuitionistic fuzzy sets from their original space to a high dimensional space （or kernel space）, so as to have shorter computational time and more accurate result. Besides, it was robust to the noises because it improved the constraint conditions used in the existing intuitionistic fuzzy clustering algorithm. Finally, compared with the traditional algorithm, the proposed algorithm has made some significant progress, and the experimental result has proved its effectiveness.