一种改进的模糊C均值聚类算法簇数量优化研究

Optimization of the Clusters Number of an Improved Fuzzy C-means Clustering Algorithm

聚类分析是无监督模式识别领域最重要的研究课题之一。模糊聚类由于建立了从样本到类别的不确定性描述,能够更客观地反映真实的世界。传统模糊聚类算法无法实现最优化配置的聚类个数自动计算。本文通过采纳分层聚类思想,提出了一种能自动、高效确定最佳聚类数目的新型自适应模糊C均值聚类算法——A-FCM算法。数值实验表明,与其他通过各种有效性函数来确定聚类数目的自适应模糊聚类算法相比,A-FCM算法的性能更优越。

Clustering analysis is one of the most important research topics in the field of unsupervised pattern recognition. Because fuzzy clustering has established the uncertainty description from samples to category,it is able to response to the real world more objectively.Traditional fuzzy clustering algorithm is unable to realize the clustering number automatic calculation with optimization configuration.Through adopting hierarchical clustering method,this article puts forward a kind of new adaptive fuzzy c-means clustering algorithm-A-FCM algorithm,which determines the best clustering number automatically and efficiently. Numerical experiments have shown that being compared to other adaptive fuzzy clustering algorithms which determine the clustering number through a variety of effectiveness functions,this method is superior in performance.