-
摘要: 提出了一种新的K值可以变化的FCM-VKNN(Fuzzy C-Means Variable K-Nearest Neighbor)聚类算法.FCM-VKNN聚类算法充分吸取了FCM算法和KNN准则的长处,使本 算法不受初始值的影响和固定值K的束缚.新的目标准则函数考虑了数据集样本的模糊隶属 关系和样本几何分布两个方面的因素,使算法的鲁棒性和分类的正确性大大加强.最后给出了 几组具有代表性数据的聚类结果.实验结果表明了这种算法的有效性.Abstract: In this paper, a new FCM-VKNN algorithm of clustering is proposed. It inherits the good virtue of FCM and KNN, which can be immune to initial guesses and get rid of the influence of constant K. Two factors are considered in cluster-validity criterion to enhance the robustness of algorithm and the validity of clustering, one is fuzzy membership and the other is geometric property. The finality of the paper gives clustering result of some representative data sets. Experimental results show the validity of the new clustering algorithm.
-
Key words:
- FCM-VKNN /
- cluster-validity function /
- fuzzy membership
计量
- 文章访问数: 2365
- HTML全文浏览量: 142
- PDF下载量: 875
- 被引次数: 0