Live streaming: Data mining and behavior analysis

Abstract

With the rapid development of mobile communication and Internet technologies, online live streaming has gradually become popular for information communication and entertainment in the new media environment. Live streaming has been widely used in teaching, reality show, E-sports games and events, brand marketing and other aspects. With the active participation of millions of streamers and hundreds of millions of viewers, massive online crowd behavior activity data are generated, which offers rich experimental scenarios for large-scale crowd behavior dynamics research, live streaming channel recommendation and online community evolution. In this paper, we summarize the relevant research literature of live streaming, and review current studies from a comprehensive list of aspects: workload pattern, viewers and streamers behavior, community network discovery and analysis, etc. We summarize the temporal and spatial patterns of live streaming platform workload, heavy tailed effect of large-scale crowd behavior in live streaming platform, etc. We believe that the future work on live streaming can be directed in the examination of formation and evolution mechanism of various community networks formed by large-scale users, as well as the recommendation and detection of live streaming content.