Influence Maximization Based on Minimum Redundancy Feature Selection in Pocket Switched Networks

Abstract

It is very crucial for viral marketing to determine a handful of influential users who can disseminate the product information to the users in a network. Influence maximization is a problem of finding users to maximize the influence to all users in networks. In Pocket Switched Networks, all users do not have information about the entire network topology. Thus, it is essential to predict the future contact opportunities for forwarding messages by analyzing the node behaviors. In this paper, we propose an influence maximization scheme based on contact frequency of the users in the network. In the scheme, we obtain the contact frequency among the users in a regular time interval during the warm-up period. At the end of the warm-up period, the main server performs the feature selection process with the contact frequency. The minimum redundancy feature selection process yields the most important features, indicating the most influential users. Finally, the server selects promising users as the influential users with the feature selection. We show that the simulation results for the diffusion time performance of the proposed scheme. The proposed scheme shows 8.5% and 12.9% more improved performance than the random and set-cover scheme.