Visualization of Graph Similarity on Irish data set

As you see, the similarity in N*N dimension can separate the groups perfectly, so after running the k-means directly on graph similarity, we can have 100 percent accuracy.