Classifying perceptions from Stakeholders about Brazil’s future using machine learning

• Autor(a) principal: da Silva, Amauri Ornellas
• Data de Publicação: 2023
• Outros Autores: Raminelli, Daniele Gonçalves de Toledo Luchetta, dos Santos, Bruno Samways, Lima, Rafael Henrique Palma
• Tipo de documento: Artigo
• Idioma: por
• Título da fonte: AtoZ (Curitiba)
• Texto Completo: https://revistas.ufpr.br/atoz/article/view/84075
• topic: Machine learning; Classification; Brazil; Stakeholders.; Engenharia de Produção: Pesquisa Operacional; Aprendizado de máquina; Classificação; Brasil; Stakeholders

This paper compares five machine learning (ML) techniques to classify the perceptions of stakeholders as to Brazil's future. The ML techniques used were artificial neural networks, k-nearest neighbors, naïve bayes, random forest and support vector machines. They were applied to a dataset retrieved from the World Bank about Brazil's development. The dataset was preprocessed and configured in two different versions: the first contained a subset of attributes manually selected by the authors, whereas the second was composed of attributes selected using the information gain approach. It was found that all ML techniques performed better using the second version of the dataset, where attributes were ranked based on information gain. However, within each version of the dataset all techniques had similar performance. This research also found that the most relevant attributes are related to business opportunities, development indexes associated with critical subjects and trust in institutions and organizations.