A probability based algorithm for influence maximization in social networks

In a social network, information runs from word-of-mouth based on the relationship of the users. The influence maximization is to find a limited number of initial users (nodes) to spread the information, so that the maximum number of other users could accept the information, which is a useful technique for marketing, information monitoring and advertising in a social network. Diffusion model of social networks imitates the process of information spreading in social networks, and Independent Cascade (IC) Model and Linear Threshold (LT) Model, are well-known stochastic information influence models. In this paper, we extend the classical IC model according to the observation of users' behaviors in social networks and propose an effective influence maximization algorithm based on this extended IC model. This novel algorithm calculates the influence probability of each node in sub-graphs that other nodes can engendered to it iteratively. The simulation experiments on real social network datasets show that our algorithm is much faster than the greedy hill-climbing algorithm, while the results are very close to the greedy algorithm and out-perform the other heuristic algorithms.