摘要：The density peak clustering algorithm(CFSFDP) is a new clustering algorithm that implements simple, clustering non-spherical data sets. The algorithm needs artificial selection of clustering center,it is difficult to get the actual clustering centers accurately and can not effectively deal with various data sets. And the density calculation processhas nonlinear time complexity. In response to the above problems, a threshold-based parallel optimization CFSFDP (PT-CFSFDP) algorithm is proposed, which sets the threshold for the local density of samples and the distance to the points with higher local density, the sample point is selected as the cluster center whenthe parameter is greater than the threshold. The distance matrix is optimized in parallel with OpenMP. Experiments show that the PT-CFSFDP algorithm can get the clustering center accurately, the accuracy of the clustering results is up to 94% and the speedup of the algorithm is up to 4.25.
