Spectral Clustering of Psychological Networks
Spectral clustering is a well-known method for clustering the vertices of an undirected network. Although its use in network psychometrics has been limited, spectral clustering has a close relationship to the commonly-used walktrap algorithm. In this paper, we report results from four simulation experiments designed to evaluate the ability of spectral clustering and the walktrap algorithm to recover underlying cluster structure in networks. The salient findings include: (1) the cluster recovery performance of the walktrap algorithm can be improved slightly by using exact K-means clustering instead of hierarchical clustering; (2) K-means and K-median clustering led to comparable recovery performance when used to cluster vertices based on the eigenvectors of Laplacian matrices in spectral clustering; (3) spectral clustering using the unnormalized Laplacian matrix generally yielded inferior cluster recovery in comparison to the other methods; (4) when the correct number of clusters was provided for the methods, spectral clustering using the normalized Laplacian matrix led to better recovery than the walktrap algorithm; (5) when the number of clusters was unknown, spectral clustering using the normalized Laplacian matrix was appreciably better than the walktrap algorithm when the clusters were equally-sized, but the two methods were competitive when the clusters were not equally-sized. Overall, both the walktrap algorithm and spectral clustering of the normalized Laplacian matrix are effective for partitioning the vertices of undirected networks, with the latter performing better in most instances.