TMsDP: two-stage density peak clustering based on multi-strategy optimization

Purpose The density peak clustering algorithm (DP) is proposed to identify cluster centers by two parameters, i.e. ρ value (local density) and δ (the distance between a point another with higher value). According the center-identifying principle of DP, potential should have than other points. However, this may limit DP from identifying some categories multi-centers or in lower-density regions. In addition, improper assignment strategy could cause wrong result for non-center This paper aims address aforementioned issues improve performance DP. Design/methodology/approach First, as many possible, authors construct point-domain introducing pinhole imaging extend searching range centers. Second, they design different novel calculation methods calculating domain distance, similarity. Third, adopt similarity achieve merging process optimize final results. Findings experimental results on analyzing 12 synthetic data sets real-world show that two-stage based multi-strategy optimization (TMsDP) outperforms state-of-the-art algorithms. Originality/value propose DP-based method, TMsDP, transform relationship points into domains ultimately further