k-Consistent Influencers in Network Data

Enliang Xu, Wynne Hsu, Mong-Li Lee, Dhaval Patel

Abstract: With the prevalence of online social media such as Facebook, Twitter and YouTube, social influence analysis has attracted considerable research interests recently. Existing works on top-k influential nodes discovery find influential users at single time point only and do not capture whether the users are consistently influential over a period of time. Finding top-k consistent influencers has many interesting applications, such as targeted marketing, recommendation, experts finding, and stock market. Identifying top-k consistent influencers is a challenging task. First, we need to dynamically compute the total influence of each user at each time point from an action log. However, to find the consistent top-scorers, we need to sort and rank them at each time point. This is computationally expensive and not scalable. In this paper, we define the consistency of a node based on its influence and volatility over time. With the help of grid index, we develop an efficient algorithm called TCI to obtain the top-k consistent influencers given a time period. We conduct extensive experiments on three real world datasets to evaluate the proposed methods. We also demonstrate the usefulness of top-k consistent influencers in identifying information sources and finding experts. The experimental results demonstrate the efficiency and effectiveness of our methods.