一种基于网格划分的密度峰值聚类改进算法

针对密度峰值聚类(Density Peak Clustering,DPC)算法具有时空复杂度高而降低了对大规模数据集聚类的有效性,以及依靠决策图人工选取聚类中心等缺点,提出基于网格的密度峰值聚类(G-DPC)算法。采用基于网格的方式进行网格划分,用网格代表点替换网格单元整体;对各代表点聚类,通过改进的自适应方法选出核心网格代表点作为聚类中心;将剩余点归类,剔除噪声点。仿真实验验证了该算法对大规模数据集和高维数据集聚类的有效性。

AN IMPROVED DENSITY PEAK CLUSTERING ALGORITHM BASED ON GRID

The high spatial complexity of density peak clustering algorithm(DPC) reduces the validity of large data clustering,and the cluster center needs to be manually selected by decision graph.To solve these problems,we proposed an improved DPC algorithm based on grid(G-DPC).This algorithm adopted the grid clustering algorithm to divide the grid and replaced the grid cell with grid representative point.Then,all grid representative points were clustered,and the core grid representative points were selected as the cluster center by the improved adaptive method.Finally,the remaining points were classified and the noise points were eliminated.Simulation results show that the algorithm is effective for large data sets and high dimensional data clustering.