Social influence analysis in large-scale networks

by Tang et al. (tang2009), SIGKDD

Tang et al. (2009) propose Topical Affinity Propagation (TAP) for determining the topic-level social influence of nodes in large networks. They introduce a distributed learning algorithm based on map reduce to compute influence graphs using TAP for identifying

• the most representative nodes for a given topic
• the influence of neighbor nodes on a particular node
• ways to quickly connect to a particular node through strong social ties

According to Tang et al. the main challenges for computing social influence graphs are:

• multi-aspect: nodes can have different influence on each other depending on the topic
• scalability
• link-specific importance (weight) between nodes rather than fixed node importance values.