Abstract
Video-based social media platforms such as YouTube and Instagram have attracted a vast number of users and subscribers in recent years. Nowadays, the production of online video clips is largely driven by individual users rather than traditional mass media. On such video-based social media platforms, many content creators collaborate with other creators to attract subscribers and diversify their content. This behavior can be seen as "coopetition" as they cooperate for their channels' success while competing with one another for the limited audience pool. This raises a question about what the coopetition networks of video creators look like and how such networks are formed. In this project, we use YouTube data on the topics of beauty and gaming to examine YouTubers' collaboration types and network structures. Specifically, we collect about 14,000 videos using keywords about beauty and gaming. We use machine learning models to filter out irrelevant videos from the raw data, then randomly sample 100 YouTube videos per topic to examine whether they are collaboration-based or not. Using the channels of the identified collaboration-based videos as the seed channels, we manually collected the guest YouTubers recursively. We model the number of collaboration videos between two channels as the weight of the edge, and the channels are modeled as the nodes of the network. The network analysis suggests that (1) the coopetition networks of YouTubers may present a scale-freeness in their topological structure and (2) beauty YouTubers collaborate with non-beauty channels more than gaming YouTubers, implying that there might be a different level of heterogeneity in collaborations depending on topics. The results inform the mechanisms of online video producers' collaboration and competition processes at an ecological level.
How to Cite:
, S., , J., , A., , K., , H., , J. & , M.,
(2022) “How Do YouTubers Collaborate With Each Other? A Machine Learning and Network Analysis-Based Approach”,
Journal of Student-Scientists' Research 4.
doi: https://doi.org/10.13021/jssr2022.3426
Files
Downloads are not available for this article.