Social media video popularity evolution analyzing based on information spreading model

Abstract

Understanding the videos' popularity evolution in social media is important to service providers, video uploaders and viewers. Such understanding can not only drive the improvement of load balancing in the network, but also can be helpful in advertising and discovering new business opportunities. While concentrating on the evolution of video's cumulative number of views after uploading, by modifying and extending an information spreading model, we characterize the popularity evolution process of videos from a new angle. In this paper, we propose a two-phase algorithm for video popularity evolution analysis: (1) For each video, we use the modified information spreading model to divide its popularity evolution process into several stages, and extract the corresponding features by fitting the parameters of model; (2) For the key feature of each stage, the increasing rate of cumulative number of views, we optimize it by using the idea of the least-squares method and improve the performance of the model. The result shows that we are able to accurately reproduce the popularity evolution process, and we also obtain some new understandings about this process.