Fuzzy U Nearest Neighbor Adaptive Clustering Algorithm

This paper introduces a fuzzy U nearest neighbor (FUNN) adaptive clustering algorithm. It initializes cluster number and cluster center based sample space density. Generally, because most experiments need to classify important clusters but not all clusters, FUNN defines U nearest neighbor concept to restrict the membership for removing noises, isolated points and uninterested data. Adding new cluster and deleting too small cluster carries out the cluster¿s life and death. So that the new algorithm is stability and the cluster accuracy is improved. Comparing with the K-Means algorithm and Fuzzy C-Means, FUNN is more effective in veracity and adaptive capability, especially processing data set included lots of noises, isolated points and uninterested data.