Descoberta de perfis de youtubers via clusterização

Abstract

Internet applications such as YouTube allow individuals to create content and post to others for viewing and interacting virtually freely and easily. Over the past few years, a wealth of user-generated information has been produced that often shares their feelings, opinions, and events in their videos. This enables the development of applications that use this information in other types of recommendation systems that are based on the content of your suggestions, and also assist in the understanding of how communication people are changing over the years. However, analyzing this large volume of data unauthorized is a nontrivial problem. Following this motivation, this paper proposes to use Machine Learning techniques and algorithms to discover and identify youtuber profiles, but specifically clustering. To do this, the video subtitles texts generated automatically by YouTube are used as the database, and then the content covered is grouped based on a similarity function. From the results, it was found that it is possible to use clustering techniques to find groups of similar youtubers in previously unknown combinations. At the end of the research, it was concluded that simple clustering techniques can be used to find groups of youtubers based on the content of your videos in a minimally satisfactory way.