Tesis doctoral de Michele Starnini
The data revolution experienced by social science has revealed complex patterns of interactions in human dynamics, such as the heterogeneity and burstiness of social contacts. The recently uncovered temporal dimension of social interactions calls for a renewed effort in analysis and modeling of empirical time-varying networks. This thesis contributes to pursue this program, through a twofold track: the modeling of dynamical social systems and the study of the impact of temporally evolving substrates on dynamical processes running on top of them. Firstly, we introduce some basic concepts and definition of time-varying networks formalism, and we present and analyze some empirical data of face-to-face interactions, discussing their main statistical properties, such as the bursty dynamics of social interactions. The main body of the exposition is then split into two parts. in the first part we focus on the modeling of social dynamics, with a twofold aim: reproduction of empirical data properties and analytic treatment of the models considered. We present and discuss the behavior of a simple model able to replicate the main statistical properties of empirical face-to-face interactions, at different levels of aggregation, such as individual, group and collective scales. The model considers individuals involved in a social gathering as performing a random walk in space, and it is based on the concept of social «attractiveness»: socially attractive people (due to their status or role in the gathering) are more likely to make people stop around them, so they start to interact. We also devote attention to the analytic study of the activitydriven model, a model aimed to capture the relation between the dynamics of time-varying networks and the topological properties of their corresponding aggregated social networks. Through a mapping to the hidden variable model, we obtained analytic expressions for both topological properties of the time-integrated networks and connectivity properties of the evolving network, as a function of the integration time and the form of the activity potential. in the second part of the thesis we study the behavior of diffusive processes taking place on temporal networks, constituted by empirical face-to-face interactions data.We first consider random walks, and thanks to different randomization strategies we introduced, we are able to single out the crucial role of temporal correlations in slowing down the random walk exploration. Then we address spreading dynamics, focusing on the case of a simple si model taking place on temporal networks, complemented by the study of the impact of different immunization strategies on the infection outbreak. We tackle in particular the effect of the length of the temporal window used to gather information in order to design the immunization strategy, finding that a limited amount of information of the contact patterns is sufficient to identify the individuals to immunize so as to maximize the effect of the vaccination protocol. our work opens interesting perspectives for further research, in particular regarding the possibility to extend the time-varying networks approach to multiplex systems, composed of several layers of interrelated networks, in which the same individuals interact between them on different layers. Empirical analysis of multiplex networks is still in its infancy, indeed, while the data mining of large, social, multi-layered systems is mature to be exploited, calling for an effort in analysis and modeling. Our understanding of the impact of the temporal dimension of networked structures on the behavior of dynamical processes running on top of them can be applied to more complex multi-layered systems, with particular attention to the effect of temporal correlation between the layers in the diffusion dynamics.
Datos académicos de la tesis doctoral «Time-varying networks approach to social dynamics: from individual to collective behavior.«
- Título de la tesis: Time-varying networks approach to social dynamics: from individual to collective behavior.
- Autor: Michele Starnini
- Universidad: Politécnica de catalunya
- Fecha de lectura de la tesis: 24/10/2014
Dirección y tribunal
- Director de la tesis
- Romualdo Pastor Satorras
- Tribunal
- Presidente del tribunal: esteban Moro egido
- claudio Castellano (vocal)
- (vocal)
- (vocal)