| Summary: | The study of infectious disease dynamics, termed epidemiology, has been an important area of research for hundreds of years. Nowadays, it is increasingly realised that such techniques may have application to epidemics of a socio-technological nature. Indeed, the proliferation of the Internet has created new opportunities to study the mechanisms behind the emergence and dynamic behaviour of online phenomena such as Internet-based popularity outbursts. The contributions of this thesis are threefold. Firstly, we explore how classical epidemiological models can be applied to model the Internet-based spreading of YouTube video views and BitTorrent downloads. We assess the potential for epidemiology to explain such phenomena, by progressively fitting and parameterising mono-epidemic models from a single data trace. We investigate the characterisation of parameter uncertainty by applying maximum likelihood-based techniques to obtain isosurfaces for different confidence intervals. We also study parameter recoverability from single stochastic simulation trajectories. Secondly, we propose a novel paradigm for modelling and predicting Internet-based phenomena. This framework is based on the composition of multiple compartmental epidemiological models, each of which is hypothesised to correspond to an underlying spreading mechanism. Our multiple-epidemic modelling approach regards data sets as the manifestation of a number of synthesised epidemics. This approach is termed 'synthedemic' modelling. It is inspired by Fourier analysis, but instead of harmonic wave forms, our components are compartmental epidemiological models. We present results from applying the synthedemic model to several epidemic outbreak datasets: synthetic SIR/SEIR, Influenza, Swine flu reported cases, YouTube video views and BitTorrent music downloads. Finally, we extend the well-known SIR model in order to investigate the potential influence of reinforcing and inhibiting interactions between epidemics. The result is the first mathematical model that can reflect the dynamics of mutually reinforcing or inhibiting epidemics, via the syndemic and counter-syndemic interaction effects in multiple overlapping populations. Our findings relating to the effect of the degree of overlap between populations are consistent with existing literature on travel restrictions.
|
|---|
