A New Density Peak Clustering Algorithm Based on Cluster Fusion Strategy

When the density peak clustering algorithm deals with complex datasets and problem of multiple peaks in same cluster, subjectively selected cluster centers are not accurate enough, allocation non-cluster is prone to joint several errors. To solve above problems, we propose a new based on fusion strategy. First, screens out candidate by setting two thresholds avoid influence noise points outliers. Second, remaining data allocated according obtain initial clusters. Third, considering structural characteristics spatial distribution datasets, definitions boundary points, inter-cluster intersection provided. correctly classify problems strategy proposed, which only corrects errors but also selects centers. Finally, test effectiveness proposed algorithm, compared DPC-KNN, DPC, K-means DBSCAN nine synthetic six real datasets. The experimental results demonstrate that performance outperforms other algorithms.