Opinion Leaders for Information Diffusion Using Graph Neural Network in Online Social Networks

Abstract

Various opportunities are available to depict different domains due to the diverse nature of social networks and researchers' insatiable. An opinion leader is a human entity or cluster of people who can redirect human assessment strategy by intellectual skills in a social network. A more comprehensive range of approaches is developed to detect opinion leaders based on network-specific and heuristic parameters. For many years, deep learning–based models have solved various real-world multifaceted, graph-based problems with high accuracy and efficiency. The Graph Neural Network (GNN) is a deep learning–based model that modernized neural networks’ efficiency by analyzing and extracting latent dependencies and confined embedding via messaging and neighborhood aggregation of data in the network. In this article, we have proposed an exclusive GNN for Opinion Leader Identification (GOLI) model utilizing the power of GNNs to categorize the opinion leaders and their impact on online social networks. In this model, we first measure the n-node neighbor's reputation of the node based on materialized trust. Next, we perform centrality conciliation instead of the input data's conventional node-embedding mechanism. We experiment with the proposed model on six different online social networks consisting of billions of users’ data to validate the model's authenticity. Finally, after training, we found the top-N opinion leaders for each dataset and analyzed how the opinion leaders are influential in information diffusion. The training-testing accuracy and error rate are also measured and compared with the other state-of-art standard Social Network Analysis (SNA) measures. We determined that the GNN-based model produced high performance concerning accuracy and precision.