Density Peak Clustering Based on Firefly Algorithm

摘要

In density peak clustering the choice of cut-off distance is not theoretically supported, and to address this concern, we propose density clustering based on firefly algorithm. The certainty between data is determined on the basis of density estimation entropy. The cut-off distance corresponding to the minimum entropy is found by iterative optimization of FA, and then substituted into the standard density clustering algorithm. Simulation experiments are conducted on eight artificial datasets. Compared with the standard density peak clustering, our method can choose the cut-off distance in a self-adaptive manner on different datasets, which improves the clustering effect.