TWITCHCOMM: A Community-Based Recommender System for Twitch Users

Twitch is currently the leading streaming platform, boasting an extensive library of content and a vast user base comprising streamers and viewers. Given this, navigating the search for content or creators to follow can be challenging. Recommender systems play a crucial role in facilitating this exploration. In this study, we developed a recommender system named TWITCHCOMM, which is designed to suggest new users with similar interests. The data used to train the recommender system was collected from SNAP. Our algorithm employs a topological approach based on linkages, specifically mutual follows, among nodes (users) to identify communities within the dataset. Subsequently, depending on whether a user is part of the dataset or an external entity, a list of recommendations is generated within their community using either a link prediction algorithm or a popularity algorithm. This recommendation algorithm is integrated into an application architecture that efficiently delivers recommendations by simply querying a Twitch account’s username.