Clustering is widely used as an unsupervised learning algorithm. However, it is often necessary to manually enter the number of clusters, and the number of clusters has a great impact on the clustering effect. At present, researchers propose some algorithms to determine the number of clusters, but the results are not very good for determining the number of clusters of data sets with complex and scattered shapes. To solve these problems, this paper proposes using the Gaussian Kernel density estimation function to determine the maximum number of clusters, use the change of center point score to get the candidate set of center points, and further use the change of the minimum distance between center points to get the number of clusters. The experiment shows the validity and practicability of the proposed algorithm.

Unsupervised clustering algorithms have been extensively applied in various research fields. In cluster analysis, determining the best cluster number k is significant. At present, many cluster validity indices (CVIs) have been proposed, including the Calinski-Harabaz index, silhouette index, and Dunn index. However, these CVIs are not effective in determining k for ring, semi-ring, and manifold data sets. We propose MPV (minimum distance and peak variation), a density-peak-based method for automated k detection. It constructs a separation measure through a dual mechanism: (1) distances between density peaks and (2) minimum distances between clusters, while using the maximum intra-cluster minimum distance to model compactness. The optimal for ring, semi-ring, and flow datasets is determined by detecting the largest change in the separation measure sequence. Experiments on 28 synthetic and real-world datasets demonstrate MPV’s accuracy in determining k for complex structures.