The spatial-temporal clustering by fast search and find of density peaks (ST-cfsfdp) has a better clustering effect on the spatiotemporal data set in a small space. However, there are some deficiencies in the spatiote...
详细信息
ISBN:
(纸本)9781665449359
The spatial-temporal clustering by fast search and find of density peaks (ST-cfsfdp) has a better clustering effect on the spatiotemporal data set in a small space. However, there are some deficiencies in the spatiotemporal dataset with large data volume and far interval between sample points, the clustering results showed great differences, too many interference points during visualization. Given the above deficiencies, this paper proposes a spatial-temporal clustering by fast search and find of density peak algorithm based on Euclidean distance constraint, by increasing the partition constraint of some sample points, the problems existing in the spatiotemporal clustering algorithm of ST-cfsfdp are improved. Experimental results show that the improved algorithm has a better clustering effect than the original algorithm.
The main idea of the recent proposed clustering by fast search and find of density peaks (cfsfdp) clustering algorithm depicts the cluster center with the local density and distance and it achieves significant effects...
详细信息
ISBN:
(纸本)9781538672556
The main idea of the recent proposed clustering by fast search and find of density peaks (cfsfdp) clustering algorithm depicts the cluster center with the local density and distance and it achieves significant effects in typical applications. But it can't select the cluster centers adaptively. Therefore, an improved cfsfdp algorithm is proposed in this paper, which determines the cluster centers by the Max-min algorithm. First, the Max-min algorithm is introduced to obtain the number of categories. Then the local density and distance information is used to determine the cluster centers as do in cfsfdp algorithm. It can not only adaptively obtain categories number of the data, but also obtain the corresponding clustering centers. The simulation results show that the proposed algorithm can find the number of categories and find the cluster centers. Meanwhile, it can find the cluster centers which are hard to be obtained through decision diagram.
暂无评论