Computer Engineering and Applications ›› 2006, Vol. 42 ›› Issue (15): 20-.

• 博士论坛 • Previous Articles     Next Articles

Research of Clustering Validity on Dot Density Weighted Fuzzy C-Means Algorithm


  1. 四川理工学院计算机科学系
  • Received:2006-02-10 Revised:1900-01-01 Online:2006-05-21 Published:2006-05-21
  • Contact: 刘小芳



  1. 四川理工学院计算机科学系
  • 通讯作者: 刘小芳

Abstract: Fuzzy C-means algorithm is an unsupervised pattern recognition method. Clustering precision of the algorithm is affected by its equal partition trend for data sets. A dot density weighted fuzzy C-means algorithm is proposed by using density size of data dot regarded as weighted value and distributing characteristic of data’s own. The method has not only to certain extent overcome limitation of fuzzy C-means algorithm, but also been favorable convergence. Clustering effect is more improved when clustering has been instructed by supervised information of a few data dots of known clustering. Clustering validity of the algorithm is evaluated by clustering validity function, thereby that has offered theoretic basic for clustering performance of the algorithm.

Key words: fuzzy C-means algorithm, weighted fuzzy C-means algorithm, clustering validity, fuzzy clustering analysis

摘要: 模糊C-均值(FCM)算法是一种非监督的模式识别方法。由于该算法具有对数据集进行等划分的趋势,影响其聚类精度。利用数据点的密度大小作为权值,借助数据本身的分布特性,提出了一种点密度加权模糊C-均值算法。该方法不仅在一定程度上克服了FCM算法的缺陷,而且具有良好的收敛性。当以聚类已知的少量数据点作为监督信息指导聚类,聚类效果进一步改善。并用聚类有效性函数对算法的聚类有效性进行了评价,从而为算法的聚类性能提供了理论依据。

关键词: 模糊C-均值算法, 加权模糊C-均值算法, 聚类有效性, 模糊聚类分析