Abstract: The traditional K-medoids clustering algorithm is sensitive to the initial center points,can’t effectively deal with dynamic data clustering,and needs privacy protection for private data.Therefore,this paper proposed the KDCK-medoids dynamic clustering algorithm.It establishes a new KD-tree using kth rectangular units optimally selected by KD-tree and incremental data based on differential privacy protection technologies,and then distributes the incremental data into the corresponding clusters by using the neighbor search strategy,and then completes the dynamic clustering.Through experiments on small data sets and multi-dimensional large data sets,clustering accuracy and running time are analyzed.And the effectiveness of the algorithm is evaluated.The experimental results indicate that the KDCK-medoids dynamic clustering based on differential privacy protection can realize privacy protection meanwhile quickly and efficiently process the dynamic clustering of incremental data problem.

Key words: KD-tree,K-medoids clustering algorithm,Differential privacy,Dynamic clustering