Data Analysis and Friendship Prediction for Twitch Streamers

Abstract

Designing recommendation systems for social networks is a common practice, and live-streaming platforms are not an exception. However, due to data and processing limitations not much work has been done to analyze these networks. In this paper, we analyzed a Twitch network gamers dataset and designed a new recommendation framework based on the specific characteristics of this dataset. The framework consists of three different layers: data, interest, and recommendation layer, each considering specific tasks. The results show the effectiveness of these friendship connection predictions among users.