Influence maximization through user interaction modeling

A majority of influence maximization models in social networks in literature are based on a seminal work by Kempe et al., in which two classic influence models were proposed i.e Linear Threshold Model and Independent Cascade Model. However, these two models use assumed values to model influence and influence propagation in social networks. This may lead to inaccurate approximation of influence. In this work, we model influence from actual social actions among members of a social network through a proposed algorithm - Selective Breadth First Traversal - that efficiently generates an optimal seed set for influence maximization. Experimental results on real data show that our approach provides an improvement over a number of traditional influence maximization algorithms.