Modelo para extração de dados e análise de sentimentos em cidades intermediárias: uma abordagem utilizando usuários centrais do Twitter

• Ano de defesa: 2023
• Autor(a) principal: De Bortoli, Arthur Facin
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Tecnológica Federal do Paraná; Pato Branco; Brasil; Programa de Pós-Graduação em Engenharia de Produção e Sistemas; UTFPR
• Palavras-chave em Português: Banco de dados; Cidades inteligentes; Algorítmos computacionais; Aprendizado do computador; Data bases; Smart cities; Computer algorithms; Machine learning; CNPQ::ENGENHARIAS::ENGENHARIA DE PRODUCAO; Engenharia/Tecnologia/Gestão
• Link de acesso: http://repositorio.utfpr.edu.br/jspui/handle/1/30878

It is believed that the global urban population, in the year 2050, will reach the mark of 68%. In the wake of the growing, rapid and continuous urbanization, management challenges have become more complex, which has made cities, thinking about providing a better quality of life for their citizens, start to adopt concepts of Smart Cities. A city becomes reasonably technological, from the point of view of Smart Cities, when it is able to make use of ICT tools to solve urban problems. As a subset of these ICT tools, there are social media, in which are produced continuous data, of wide diffusion and dissemination. The ability to know what subject citizens are talking about the most, and how they feel about it, through data from social media, is something that meets the use of ICT, aimed at the context of Smart Cities. However, the task of extracting and analyzing data from social media is not trivial, especially in Brazilian cities of intermediate size, which calls for new solutions. Thus, what is proposed here is a model for extracting and analyzing data, having the social media Twitter as the source of content, geographically located in the context of intermediate cities, with the city of Maringá as the locus of this study. For that, an approach of central users is used, in the social network Twitter, to extract the data. For analysis, Dirichlet Latent Allocation (LDA) is applied to identify the prominent topic(s), and Sentiment Analysis to discover its polarity. The data extraction located in the context of cities, using both the central users approach and the geotagged approach was successful as was the application of the LDA, indicating the most proeminent topics, which made it possible to verify the themes most discussed by netizens. The main topics identified through the LDA were related to: Pandemic, with data collected in January 2022, in the first model test, and Urban Mobility, with data collected from September to December 2022, in the second model test. Carrying out the thematic collection on these subjects, the predominant sentiment within the data was negative. It was also found that the supervised strategy was more successful for classifying the sentiment polarity of tweets, especially the SVM and Random Forest methods, especially in the second application test of the model. Having the contribution of the literature and subsequent performance of two tests, the final flowchart as well as a generic model are presented at the end, being, subsequently, raised limitations of the work and opportunities for future studies.