An Improved Density Peaks-Based Graph Clustering Algorithm

The density peaks algorithm is a widely accepted density-based clustering algorithm, which shows excellent performance for many discrete data with any shape, and any distribution. However, because the traditional node density and density following distance does not match the graph data, the traditional density peaks model cannot be directly applied to graph data. To solve this problem, an improved density peaks graph clustering algorithm is proposed, simply called DPGC. Firstly, a novel node density is defined for the graph data based on the aggregation of the relative neighbors with a fixed number. Secondly, a density following distance search method is designed for graph data to calculate the density following distance of each node, so as to enhance the accuracy of selecting cluster centers. Finally, an improved density peaks model is constructed to quickly and accurately cluster the complex network. Experiments on multiple synthetic networks and real networks show that our algorithm offers better graph clustering results.