Increasing the Predictive Power of Affiliation Networks

Scale is often an issue when attempting to understand and analyze large social networks. As the size of the network increases, it is harder to make sense of the network, and it is computationally costly to manipulate and maintain. Here we investigate methods for pruning social networks by determining the most relevant relationships in a social network. We measure importance in terms of predictive accuracy on a set of target attributes of social network groups. Our goal is to create a pruned network that models the most informative affiliations and relationships. We present methods for pruning networks based on both structural properties and descriptive attributes. These pruning approaches can be used to decrease the expense of constructing social networks for analysis by reducing the number of relationships that need to be investigated and as a data reduction approach for approximating larger graphs or visualizing large graphs. We demonstrate our method on a network of NASDAQ and NYSE business executives and on a bibliographic network describing publications and authors and show that structural and descriptive pruning increase the predictive power of affiliation networks when compared to random pruning.