模糊C-均值聚类算法的改进

Improvement of fuzzy C-means clustering algorithm

针对传统的模糊C-均值算法FCM受初始聚类中心影响而易于收敛到局部极小值的问题,提出了具体的改进方法。初始聚类中心不再随机获取而是通过改进的算法有目的地进行选取,同时采用冗余聚类中心的方法先将大簇分割成多个小类,再按一定条件将相邻的小类合并。实验结果表明,改进后的FCM算法减小了对初始聚类中心的依赖,聚类结果更加精确。

The traditional Fuzzy C-means algorithm has the shortage that be sensitive to the initial cluster centers, easily converge to a local minimum result. To solve the above problem, this paper presents an improved algorithm. Through the improved algorithm, the initial cluster centers are purposefully selected but not randomly selected. At the same time we can express a big cluster used several small clusters, then merger adjacent clusters that satisfy certain conditions. Experiment result demonstrates that the improved FCM algorithm can decrease the dependence on the initial cluster centers and get more accurate clustering results.