Fig. 1
From: Dynamic correlation network analysis of financial asset returns with network clustering
Network clustering by recursive hierarchical network division. Note: The whole market includes all 1,324 stocks as a single group portfolio. The network is divided into several sub-networks (marked by circles) so as to maximize the modularity of the network. The two major categories: Cyclical and Defensive groups are created at the first division of the whole market. This modularity based network division is applied to the sub-networks recursively until the final 14 groups (marked by squares) are created. Further details of the recursive division algorithm is described in Isogai (2014)