Leveraging Deep Learning to Spot Communities for Influence Maximization in Social Networks

Abstract

Groups play a crucial role in affecting decisions of individuals who are part of the group. When it comes to social networks the group here may be small with some 10-15 members or very big contacting more than 100 members. Thus, there is high possibility of individuals belonging to one or more groups in social networks. It thus becomes important to activate influential members of a group to ensure maximum information propagation. This work proposes a community-based seed selection algorithm. The communities are first identified node embedding which performs graph clustering. After which proportionate distribution of seed nodes is carried out to ensure fair selection. Mapping node features to lower dimensional space and similar nodes getting placed closer to each other proves a better technique for community detection and is also expandable if new nodes get introduced in the network.