K-NNDP: K-means algorithm based on nearest neighbor density peak optimization and outlier removal

Jiyong Liao, Xingjiao Wu, Yaxin Wu, Juelin Shu

Abstract: Highlights•K-NNDP algorithm address the combined problem of selecting initial centroids and detecting outliers.•Automatically selecting initial cluster centers using nearest neighbor density peaks.•Replacing mean with median to avoid the influence of outliers on clustering results.•Our algorithm outperforms the state-of-the-art improved K-means algorithm.