Hits:

Indexed by:会议论文

Date of Publication:2014-01-01

Included Journals:CPCI-S

Key Words:Clustering methods; Clustering algorithms; Multi-layer Multi-center; Data mining; Machine Learning

Abstract:Aiming at existing partition clustering algorithms restricted by a single clustering and separation degree depends on the initial cluster centers, For example, K-means algorithm, the K-center, etc. However the hierarchical clustering algorithm (AGNES) whose time complexity and space complexity is higher is not suitable for large-scale numerical data calculation, and the density clustering algorithms such as DBSCAN algorithm depends on the number of data points in the field of fixed radius and the threshold. It is also very sensitive to the parameters. An agglomerative clustering algorithm based on multi center atom sets is proposed, MMACA for short. This algorithm is based on the idea of multi center, in accordance with the initial number of atoms randomly forming atomic set, then removing the local noise atomic concentration, constituting the original atomic set. Finally, condensation according to changes in the radius of the original atomic nucleus set, in order to control the number of iterations of the aggregation process. The MMACA algorithm is applied to large data sets and through a lot of experiments to fully verify the reliability and validity of MMACA algorithm.