Adaptive Density Peak Clustering Based on Dimension-Free and Reverse K-Nearest Neighbours

Cluster analysis is a crucial component in consumer behaviour segmentation. The density peak clustering algorithm (DPC) is a novel density-based clustering method, but it performs poorly in high-dimension datasets and local density for boundary points. In addition, the DPC fault tolerance is affected by the one-step allocation strategy. To overcome these disadvantages, an adaptive density peak clustering algorithm based on dimension-free and reverse k-nearest neighbours (ERK-DPC) is proposed in this paper. First, we compute the Euler cosine distance to obtain the similarity of sample points in high-dimension datasets. Second, the adaptive local density formula is used to measure the local density of each point. Finally, the reverse k-nearest neighbour approach is added onto the two-step allocation strategy, which assigns the remaining points accurately and effectively. The proposed clustering algorithm was applied in experiments on several benchmark datasets and real-world datasets. After comparing the benchmarks, the results demonstrate that the ERK-DPC algorithm is superior to selected state-of-the-art methods.