Popularity prediction of live streaming content with user response

Abstract

Online live streaming is rapidly growing as the most focused content for internet users. Twitch.tv, which started as a gameplay live broadcast service, is the world’s largest internet broadcast platform with the system for viewers to interact with internet broadcasters through responses such as chatting. In this study, we collected chat data from 2162 videos of 52 channels of Twitch.tv to analyze user reactions appearing in the internet broadcast platform. Based on the user reaction data, we constructed the machine learning based video popularity prediction model and obtained high prediction performances. The user response data collected through this study can be used for future research on internet broadcasting platform and text analysis.