Methodology
Figure 1. An example of the variation of the (a) stored flow and T (s) (b) and the acc during execution of the main process of FlowPro [2].
Figure 2. The quantity S(x) (stored flow of a node) and
the local(x)/degree(x) for a synthetic graph using blue and red colors for the
nodes that
belong on the community and does not belong on community, respectively [2].
-
In each iteration of the main process of FlowPro, the initial node propagates a flow that is shared to its neighbors. Each node is able to store, propagate to its neighbors and return a part of flow to the initial node.
-
When the algorithm converges, the stored flow of the nodes that belong on the community of initial node is generally higher than the stored flow of the rest graph nodes resulting the requested community.
-
In addition, using the stored flow of the nodes, FlowPro is able to extend the neighbors of initial node decreasing shortest paths between the nodes that belong on the community and to remove the neighbors of initial node, subtracting bridges of the community.
Figure 3. The evolution of FlowPro community detection result (before the black line) in a synthetic graph [2].