Online stream clustering using density and affinity propagation algorithm

In order to improve the data stream clustering accuracy and effectiveness, the paper proposes an efficient data stream clustering algorithm based on affinity propagation and density methods (APDenStream). The algorithm adopts online/offline two stage process framework, by using micro-cluster decay density to reflect the evolution information and using online dynamic delete mechanism to maintain micro-cluster, it makes the algorithm model more consistent with the intrinsic data stream characteristics, simultaneously, The algorithm also takes advantage of the WAP algorithm which detects new class patterns merged into the clustering model. The experimental results show that this algorithm has good applicability, efficiency and extensibility and can get better clustering effect.