A Novel Approach to Predicting Live-Stream Subscribers

Live streaming is a new type of social media that uses streaming media technologies for broadcasting video in real time. The rise of live streaming services has changed consumer behaviors and brought up a new business model. Usually, live streamer is able to receive a constant income from commissions of user subscriptions and earn extra money from product sales and user donations. Prior studies focused on how to acquire viewers; while they failed to explore how to transform the viewers watching for free into valuable consumers capable of purchasing, subscribing, and even donating to streamers. This study explored the feasibility of finding potential subscribers. A predictive model for potential subscriber is built by the machine learning approach based on users' personal information and their online interaction with streamers. The preliminary result shows the predictive model constructed by the random-forest technique using daily interaction data can better predict the potential subscribers.