关键词: 关键词: 密度峰值, 网格粒化, 大规模数据, 聚类

Abstract: Abstract: The density peak clustering algorithm (DPC) proposed in 2014, possess a concise and novel idea, which requires a fewer parameters and no iteration. In this paper, a grid clustering algorithm is proposed based on DPC. First, the N dimensional space is divided into disjoint rectangular unit. Then the central cells of space is found based on the idea of DPC, namely, the central cells are surrounded by other grid cells of low local density, and the distance with grid cells of high local density is relatively large. Finally, the grid cells similar to their central cells are merged to obtain the clustering results. The experiments show that this algorithm can quickly find the clustering center, and effectively deal with the clustering problem of large scale data, which processes a higher efficiency compared with the original density clustering algorithm on different data sets, reducing the loss of time 10~100 times, and maintaining the loss of accuracy at 5% ~ 8%.

Key words: Keywords: Density peak, grid granulation, large scale data, clustering