||In recent years, there has been a new type of media, “live stream”. Through streaming technology, it can make people show their lives to others and interact with others at low latency. This kind of media greatly improve the user experience between communicators and audience. Among many live stream platforms, Twitch is most famous in Taiwan. In 2018, there were an average of 15 million people watching live stream on Twtich every day . A large number of people will bring a lot of business opportunities.|
Therefore, in this study, we want to explore the relationship between the live streamers and viewers on the Twitch platform. We want to predict whether viewers will be willing to pay for subscriptions. This study uses Twitch API and web crawler to collect data. We use text mining to catch viewers’ actions in live stream’s chat room and label general viewers and subscribers. And then, we use some supervised learning methods, such as logistic regression, SVM, decision tree and random forest, to build models to predict.
The best model’s accuracy could reach 0.73664. And the result indicates that the chat frequency, live stream’s notification and the number of viewer’s following influence whether the viewer will subscribe the streamer.