Influence Maximization Across Partially Aligned Heterogenous Social Networks

The influence maximization problem aims at finding a subset of seed users who can maximize the spread of influence in online social networks (OSNs). Existing works mostly focus on one single homogenous network. However, in the real world, OSNs (1) are usually heterogeneous, via which users can influence each others in multiple channels; and (2) share common users, via whom information could propagate across networks. In this paper, for the first time we study the influence maximization problem in multiple partially aligned heterogenous OSNs. A new model, multi-aligned multi-relational network influence maximizer (M&M), is proposed to address this problem. M&M extracts multi-aligned multirelational networks (MMNs) from aligned heterogeneous OSNs based on a set of inter and intra network social meta paths. Besides, M&M extends traditional linear threshold (LT) model to depict the information diffusion across MMNs. In addition, M&M, which selects seed users greedily, is proved to achieve a (1 - 1/e)-approximation of the optimal solution. Extensive experiments conducted on two real-world partially aligned heterogeneous OSNs demonstrate its effectiveness.