On the Badge Aware Influence Representation Learning & Influencer Detection in Social Networks

Influence maximization is a widely studied topic in social networks, where the aim is to reach the maximum population, while recommending only a small seed of influential members. Close inspection reveals that all the recommended influential members may not be equally effective in influencing people of various topics. Traditional influence maximisation techniques which depends on heuristics to find influential members suffers from poor quality of influencers. Addressing this issue, we develop Badge2vec which aims to learn the topic aware embedding of the influencers and followers, and subsequently recommend top k influencers Ik and top r influence badges Rr such that Ik becomes most effective for events hosted on topics in Rr, to optimize the total influenced population. Badge2vec is being evaluated on Meetup and Yelp. Experimental results demonstrates that (i) Badge2vec outperforms the state of the art algorithms in terms of the total influenced population, (ii) the influenced population of Badge2vec are more topically cohesive and aligned to the interest of the Meetup group in consideration.