BetClose Approach to Improve the Diffusion in Social Networks

Abstract

In the era of Internet and advancing in communication and computing technology, the social networks have caught wide applications in real life's. Influence Maximization (IM) in social networks is intended to find a small group (sub set) of specific users as seeds (influencers). These seeds have an ability to activate maximum possible neighbors in the network. IM is usually categorized as NP-hard problem. Different alternatives were proposed to develop approaches to resolve it. Most of the available approaches are considering the centrality metrics to categorize the seed nodes (initial influencer node / early adopters). The essential problem in this study is how to discover the significant players (a subset of seed nodes) in social networks in an efficient manner. Those players must have an ability to affect the maximum probable neighbors by encouraging them to adopt the diffused idea. Best selection will lead to maximize the influence propagation in the network. Information is spreading stochastically in such networks. All These steps are to maximize the information dissemination in social networks. A new proposed approach to find the minimum number of nodes that can maximize information diffusion most effectively is introduced and named as (BetClose) in this study. BetClose has an ability to assign set of nodes with highest betweenness and closeness values in the same time. A developed diffusion model (depend on ICM and LTM simultaneously) is implemented. Certain approaches are used to develop probabilities to improve the diffusion process and a mathematical model is suggested to estimate the threshold values based on the network structure. The results of this study offers indications about the structure of network that having a probable effect on the information diffusion in social networks. The suggested approach outcomes indicate high adoption compared with the traditional approaches which are depending on one metric only. The average adoption size resulted from BetClose is found to be better than the average adoption size resulted from implementing each of the three essential traditional centrality metrics (i.e. betweenness, closeness, and Eigen vector).