Data-Driven Discovery of Social Network Dynamics

In recent years, rigorous mathematical frameworks have been developed for modelling complex networks such as social networks, which can be used to determine several of their properties such as the resilience of the network to external perturbation and the propagation time of signals within them. Several modern algorithms have been proposed in order to identify models of dynamical systems from big data, such as the well-known “sparse identification of nonlinear dynamics (SINDy)” algorithm. We modify this algorithm such that given data regarding the dynamics of a social network, the differential equation that best describes the underlying dynamics of the social network is identified in accordance with the aforementioned mathematical frameworks. Due to the massive growth of the activity within social networks, the efficiency and speed of such algorithms are becoming increasingly crucial. Testing the proposed algorithm on empirical data verifies the accuracy and efficiency of our approach.