Protótipo para mineração de opinião em redes sociais: estudo de casos selecionados usando o twitter

• Autor(a) principal: Santos, Leandro Matioli
• Data de Publicação: 2015
• Tipo de documento: Trabalho de conclusão de curso
• Idioma: por
• Título da fonte: Repositório Institucional da UFLA
• Texto Completo: http://repositorio.ufla.br/jspui/handle/1/5190
• topic: Mineração de opinião; Análise de sentimento; Web mining; Aprendizagem de máquina; Classificação; Opinion mining; Sentiment analysis; Machine learning; Classification

This work aimed the development of a prototype that is able to perform opinion mining in texts from social networks, using Twitter as a study case. A detailed description of the concepts associated with the field of web mining and the process of text mining was presented, showing the current techniques for its realization. The programming language Java was used to build the prototype, but previously, a small UML modeling was done. The machine learning method known as SVM was chosen to perform the binary classification between positive and negative sentiments, which represents an action of opinion mining or sentiment analysis. To represent the text documents in a structured manner, it was opted to use the feature vector, which is a statistical approach to text analysis. In the tests executed, it was possible to observe that the trained classifier achieved an average accuracy of 80% on the desired classifications. It was concluded that success was obtained with the prototype created, because it was able to perform the tasks solicited in a reasonable way.