The spatial structure of mobile communication networks

There has been a recent surge of interest in the relationship between the spatial and topological structure of communication networks with the availability of large scale anonymous datasets on the communication and mobility patterns of individuals. These datasets, captured as a by-product of modern communications technology, provide a detailed view of the daily interpersonal interactions of millions of people. Mobile phone call logs in particular offer an unparalleled source of information given their personal portable nature and ubiquity in modern society. The use of mobile phones has become so common that these datasets are no longer merely communication logs but close approximations of the network of interpersonal relationships that forms society. The analysis of these proxy networks has the potential to uncover knowledge about society at a scale never previously possible. Networks, and social networks in particular, have been the subject of investigation for more than a century with a rich corpus of theory and methods now available to researchers. Computational approaches to the study of networks are more recent but there are now a wide variety of structural analysis methods that have been developed and applied across many different disciplines and subject areas. The study of interactions across space has developed in parallel with theory, methods, models and a variety of applications. Recent studies of these proxy networks have tended to use computational approaches for analysing community structure and modelling spatial interacitions without much regard for the theory upon which they were built. The underlying assumption has been that all phenomena that can be represented as networks can be analysed with the same methods. In this thesis we demonstrate that this is not the case and identify a number of problems and misinterpretations that can arise when inappropriate methods or network representations are employed. Through a detailed theoretical and empirical analysis we identify appropriate combinations of network representation, spatial scale, and analysis methods for studying the spatial structure of communication networks. Using these findings we demonstrate the potential of such analysis when the appropriate methodology is employed