Using Random Forests and Decision Trees to Predict Viewing Game Live Streaming via Viewers’ Comments

Abstract

In recent years, live streaming has developed rapidly in the world and become one of the most popular entertainment activities of most people since 2011, especially the youth due to the rich and various content. Previous literatures mainly focused on finding popular streamer and behaviors of live streaming viewers like gift giving behaviors. However, the studies on the effect of reviewers’ comments on the number of viewing and on text comments via live chat rooms of social media users, especially of games live streaming users are very limited, although these issues are considered to significantly and positively affect others’ behaviors. Therefore, this work aims to use the text comments in the live chat room as input variables to predict the number of viewing games live streaming. Random forests (RF) and decision trees (DT) algorithms were employed to build prediction models. Game live streaming was our research target. The prediction accuracy rate of the established model is nearly 90%. The analysis results is expected to be a roadmap for the live streaming platforms to carefully respond viewers’ comments.