An improved evolutionary density peak clustering algorithm

Aiming at the problem that traditional density peak clustering algorithm can't adapt to discover knowledge in the database with dynamic growth of data volume, in this paper, we propose an improved evolutionary density peak clustering algorithm (EDPC). The EDPC algorithm does not need any prior knowledge, and can select the control parameters according to the comprehensive index
$\gamma$
of the sample itself, which overcomes the defect that there is no reference in the selection of prior parameters in the current incremental algorithm. EDPC algorithm is based on incremental data, which effectively solves the problem that traditional DPC algorithm can not adapt to the dynamic change of data. At the same time, with the addition of new samples, EDPC algorithm can accurately find the cluster center, and adjust the control parameters according to the generated
$\gamma$
graph, so as to decide whether to generate new clusters or merge old clusters. Finally, experiments on UCI and Synthetic datasets show that EDPC algorithm is able to outperform incremental K-means and incremental DBSCAN algorithm.