Influence maximization in social networks using effective community detection

Abstract

Influence maximization problem aims to find a set of nodes with the highest diffusion in social networks in order to maximize diffusion in the graph by this set. A set of these nodes can be used to diffuse news, viruses , marketing and etc. To solve this problem, some algorithms have been proposed to help identify a set of nodes. Due to the low accuracy and high run time in selecting the set of nodes by the proposed algorithms, further studies should be performed in this area. This paper intends to solve the Richclub problem in selecting nodes, reduce the search space to decrease computational overhead, and achieve increases the algorithm accuracy by selecting high-charisma nodes. The Charismatic Transmission in Influence Maximization (CTIM) algorithm reduces the computational overhead by using community structure and pruning criteria. In this algorithm, seed nodes are selected using nodes that have high charismatic power. Also, these nodes have a high correlation with influential nodes in other communities, which causes optimal diffusion in social networks. As a result a set of seed nodes are selected by calculating the influence spread, which increases the algorithm accuracy. In all size ranges of seed sets and different data sets, experiment results show The CTIM algorithm performs well in influence spread. Also in The douban dataset significantly outperforms the PHG algorithm to as much as a 6.56% increase in influence spread and the execution time has decreased by 99%.