关键词: 聚类, 相似度, 相对互连度, 相对接近度

Abstract: In order to obtain better clustering results，this paper proposes a new hybrid clustering algorithm based on traditional partitioning and hierarchical methods（shorted for MPH）.The algorithm divides the clustering process into two phases：Splitting and merging.During the splitting process，MPH divides the dataset into a number of sub-clusters by repeatedly using k-means algorithm；and during the merging process，it clusters by agglomerative hierarchical methods.The experimental results show that the algorithm is significantly effective in discovering clusters of arbitrary shapes，sizes，and it is not sensitive to the noises.

Key words: clustering, similarity, relative inter-connectivity, relative closeness