Abstract: A decision cluster classifier based on agglomerative K-means was proposed. Combining with the cluster validation techniques, the agglomerative strategy and clustering cluster parameters WTBXλ WTBZwere applied to K-means clustering. On the training data set, the tree was established by the topdown nested clustering, then the decision classification model was extracted from the nested tree. The experimental results based on UCI  showed  that the proposed classification method had the superiorities as follows:(1) the problem of which the K-means is sensitive to the initial center position can effectively improved;(2)The number of clusters can automatically determined;(3)The density of clustering clusters can effectively controled.