Inferência em redes aleatórias com pesos discretos

• Ano de defesa: 2023
• Autor(a) principal: Costa, Laila Letícia da Silva
• Orientador(a): Cerqueira, Andressa
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de São Carlos; Câmpus São Carlos
• Programa de Pós-Graduação: Programa Interinstitucional de Pós-Graduação em Estatística - PIPGEs
• Palavras-chave em Português: Redes aleatórias; Detecção de comunidades; Modelo estocásticos em blocos; Distribuição de Poisson inflada de zeros; EM-Variacional
• Palavras-chave em Inglês: Random network; Community detection; Stochastic block model; Zero-inflated Poisson distribution; Variational EM
• Área do conhecimento CNPq: CIENCIAS EXATAS E DA TERRA::PROBABILIDADE E ESTATISTICA
• Link de acesso: https://repositorio.ufscar.br/handle/ufscar/18097

Random networks have been widely used to describe interactions between objects, including interpersonal relationships between individuals. One of the most important features of networks is the presence of communities, which are groups of nodes with similar patterns of connection. In this regard, we propose a model in which edges between pairs of vertices are randomly assigned, given the communities of those vertices, following the zero-inflated Poisson (ZIP) distribution. This proposal allows us to model networks with community structure that are sparse and have edge weights. The estimation of the parameters of the ZIP distribution is performed using the EM algorithm, while the estimation of communities is done using the EM-Variational algorithm. The performance of the estimators is evaluated through simulation studies, using the Normalized Mutual Information (NMI) comparison measure to compare the true and estimated communities. To compare the estimated parameters of the ZIP distribution, we use the Mean Squared Error (MSE). Finally, we apply the proposed model to airport networks in Brazil and detect the community structure from 2018 to 2021, in order to evaluate the changes that occurred in these networks before and during the COVID-19 pandemic period.