
187Extending Classification ofSocial Networks throughIndirect Friendships
we see here that by choosing additional edges in the graph well, we can improve the classication
accuracy of the network classier.
16.6 SUMMARY AND DIRECTIONS
We have shown here that by intelligently choosing important nodes in a subgraph, one is able to
increase classication accuracy on real-world social network data. We present a method of selecting
nodes by using weighted graph metrics and a programmatically determinable threshold parameter.
Furthermore, we show that our τ threshold is a useful tool for determining the general increase in
graph size versus a desire ...