Research Of Influence Maximization Problem Base On Targeting Social Groups

Viral marketing is one of the most important application in social network. In real life scenarios, social network users typically belong to a social group with specific organizational structure. So how to choose a given number of groups, based on mutual influence spread between nodes in all groups, so that the largest number of network users receive information, is of great importance for product marketing while making full use of social propagation characteristics of the network.For the influence maximization problem based on social groups, it 's more scalable to simulate the influence spread process at the level of social groups. We designed GLISM model to simulate the influence spread between social groups. The GLISM model include: network graph based on groups, the influence spread rules to describe influence spread upon the network graph. it also propose a method to compute the influence spread within GLISM model and discuss the related properties of it.From the perspective of greedy, design GLISMGreedy greedy algorithm Based on GLISM model, GLISMGreedy algorithm uses GLISM model to calculate the influence spread marginal gain, and combine the submodular properties of GLISM model to make the algorithm more efficient. In order to obtain more scalable algorithms, we designed the IR-DU algorithm based on group network graph constructed by GLISM model. IR-DU algorithm use the group influence as the estimate of group influence spread marginal gain and the method to compute the group influence is provided. IR-DU algorithm also use the group influence as a criterion for selecting seed groups and re-estimate group influence during the process of selecting seed groups to reduce the influence spread overlap between groups.Finally we analyzed the proposed algorithm through experiment. The experiment choose dataset DBLP and dataset Net HEPT as dataset resources, choose the basic greedy algorithm and degree based heuristic algorithm to compare from the aspect of influence spread and time efficiency. Besides we obtained and analyzed the experiment results, the experiment results show that the experimental effect of GLISMGreedy algorithm and IR-DU algorithm is very close to the basic greedy algorithm, but their efficiency of the algorithm effectively enhanced.The effect of GLISMGreedy algorithm is slightly better than IR-DU algorithm,but IR-DU algorithm if more efficient.