Semi-supervised Affinity Propagation Clustering Algorithm Based on Stratified Combination

Considering the complexity and the accuracy, an improved affinity propagation clustering algorithm Semi-supervised Affinity Propagation clustering algorithm based on Stratified Combination (SAP-SC) is proposed. In order to make the operation simplified and easily-implemented, the proposed algorithm introduces a stratified clustering method which equally partitions the integrative clustering process into several smaller blocks. Focusing on the hard clustering data, every layer employs semi-supervised learning to conceive pair-wise constraints and maps each sub-cluster with the corresponding label. Also, assembled boosting method is utilized to weight together all layered results to improve the clustering performance. Finally, theoretical analysis and experimental results show that the algorithm can achieve both higher accuracy and better computational performance.