Investigando a Associação de Canais Financeiros do YouTube na composição de carteiras: um estudo com análise de sentimentos

Abstract

YouTube has become an increasingly relevant source of information for investors, expanding the influence of financial content creators on investment-related perceptions and decisions. In this context, this study investigates potential relationships between the sentiments expressed by financial influencers in YouTube videos, the composition of the Carteira Valor, and the behavior of stocks traded on the Brazilian stock exchange. Natural Language Processing (NLP) techniques were applied to subtitle data extracted from videos of two major financial education channels. A sentiment classification model based on the Random Forest algorithm was developed, together with data balancing techniques. In the best experimental scenario, using two classes, the modelachieved an accuracy of 81%. The results indicate that, although no significant patterns were found between the extracted sentiments and the composition of the Carteira Valor, there is evidence of an association between stock mentions in the videos and short-term price variations of the respective assets.