Mining multi-layered networks - applications to the portuguese urban system and EU domains

Abstract

The main purpose of this thesis is to emphasise the importance of a multi-layered approach to a network analysis, by applying the most appropriate metrics to two case studies: a territorial network analysis capturing several insights of the Portuguese society and a study identifying the most important and influential leading countries of the European Union. On the one hand, by giving the basis for monolayer and multilayer network analysis, the researchers develop a solid selection of the main concepts and analytical measures, bearing in mind the general structure of different types of networks. On the other hand, these frameworks are tested in two real multilayer networks, showing the validity and the meaningfulness of the introduced measures, as important and non-random information to mine complex phenomena. The researchers analyse all the 308 Portuguese municipalities in some economic features that describe a society (work and education) and the current EU-28 countries composition in important priorities of union policies such as commercial trade, foreign direct investment and migration and remittance flows. For both national and European schemes, after the analysis of each variable, where communities are detected individually, an adequate method is implemented to detect communities under the perspective of a multilayer network. Subsequently, the multilayer outcomes are compared with the monolayer ones. The application of a multi-layered network allows inferring spatial patterns in a way more consistent in comparison to the geographical structure deduced from each network layer taken separately. Thus, the research problem underlying this dissertation contributes to a structured analysis model to identify the appropriate analytical measures to derive the Portuguese urban system and the European Union global structures.