Cost Effective Influence Maximisation

In the context of virality prediction, many researchers have leveraged the existence of a coreperiphery structure in a network to identify the superspreaders of information. Topologically, the nodes in the core of a network are the most efficient spreaders. However, these nodes are less susceptible, i.e., unlikely to be influenced by the periphery nodes. Consequently, large payoffs are required to market information (ideas, products, memes, etc.) via them. In this paper, we show the presence of several non-core nodes whose spreading power is close to that of the core nodes. In the 4 real-world datasets under consideration, the number of such nodes is 7 times more than the number of core nodes on average. Given a limited budget, digital marketers can target such non-core nodes to advertise their products with lesser payoffs. Moreover, from a sociological perspective, interpersonal closeness can help in reducing the payoff further. With the help of friendship connections, we propose a cost-effective strategy to reach the influential nodes. The proposed hill-climbing based strategy can be effectively used with both, global as well as local characteristics of the nodes in a network. In terms of the cost metric, it outperforms the conventional independent cascade model by more than 5 times for the core and 2 times for the non-core super-spreaders.