Social Influence Maximization Using Genetic Algorithm With Dynamic Probabilities

The previous decades have observed the exponential growth of online social networks, where billions of users exchange information with each other and generate tremendously large quantity of the content. This dominance of social networks in our daily life has encouraged more consideration of researcher in the field of information diffusion, where a small bit of information could widespread through "world of mouth" effect. One of the key research problems in information diffusion is influence maximization, which is a NP-hard problem. Influence Maximization (IM) is the problem to find k number of nodes that are most influential nodes of the network, which can maximize the information propagation in the network. Various heuristics available to find most influencing nodes of the network include random, high degree, single discount, general greedy and genetic algorithm with weighted cascade etc. In this paper, we proposed dynamic probability based genetic approach using topic affinity propagation (TAP) method to find the optimal set of influential nodes of the network. The efficiency of the proposed approach is analyzed on two large-scale networks. Results express that the proposed algorithm is able to improve the influence spread by 6% to 13% with respect to various influence maximization heuristics.