A novel density peaks clustering algorithm based on k nearest neighbors for improving assignment process

Density Peaks Clustering (DPC) algorithm is a kind of density-based clustering approach, which can quickly search and find density peaks. However, DPC has deficiency in assignment process, which is likely to trigger domino effect. Especially, it cannot process some non-spherical data sets such as Spiral. The research results indicate that assignment process appears to be the most significant step in deciding the success of the clustering performance. Therefore, we propose a density peaks clustering based on k nearest neighbors (DPC-KNN) which aims to overcome the weakness of DPC. The proposed DPC-KNN integrates the idea of k nearest neighbors into the distance computation and assignment process, which is more reasonable. It can be seen from experimental results that the DPC-KNN algorithm is more feasible and effective, compared with K-means, DBSCAN and DPC.