Clustering by Creating a Graph

In this paper, we presented a novel graph-based clustering algorithm (GC). GC contains two main steps: the first step is to create a graph and find out the key nodes as centers, the second step is to divide every data point to each center. The centers are selected from a graph view. Experimental results on 8 datasets demonstrated that GC could do better than k-means, k-medoids, Hierarchical Clustering and Gaussian Mixture Models. Moreover, the most important parameter of clustering algorithms is the number of clusters K and the comparing algorithms need K set to true number of clusters. But for GC, the only requirement is that K is not less than the true number of clusters.