基于K近邻隶属度的聚类算法研究

Clustering algorithm based on membership degree of K-nearest neighbor

经典模糊C均值聚类算法(FCM)基于欧氏距离,存在不同规模类簇不能正确聚类问题,针对此问题提出一种基于K近邻隶属度的模糊C均值聚类算法(KNN_FCM)。讨论了基于K近邻隶属度的粗糙C均值聚类算法(KNN_RCM)和粗糙模糊C均值聚类算法(KNN_RFCM),此方法避免了传统粗糙C均值聚类算法(RCM)和粗糙模糊C均值聚类算法(RFCM)中阈值选择问题。将KNN_FCM、KNN_RCM、KNN_RFCM分别与FCM、RFM、RFCM在UCI数据集上进行仿真比较,结果表明新方法是可行、有效的。

Classic fuzzy c-means algorithm（FCM）is based on Euclidean distance, it includes a problem that different size of class cluster is not clustered correctly. Aiming at the problem, this paper presents a fuzzy C-means algorithm based on the membership degree of K-nearest neighbor（KNN_FCM）. Then the paper discusses rough C-means clustering algorithm and rough fuzzy C-means clustering algorithm, which are both based on the membership degree of K-nearest neighbor.These algorithms avoide a question of threshold selection in traditional rough C-means clustering algorithm and rough fuzzy C-means clustering algorithm. Compare the KNN_FCM、KNN_RCM、KNN_RFCM with FCM、RFM、RFCM in UCI dataset, the experimental result shows that the method is feasible and effective.