Efficient estimation with Bayesian EM and inlabru methods for spatio-temporal Hawkes point processes: an application to real-world data
We present a fast and accurate method for modelling spatio-temporal Hawkes point processes, combining a Bayesian Expectation-Maximization (EM) approach with the inlabru methodology. Hawkes processes, used to model the dynamics of self-exciting events in time and space, are frequently applied to complex phenomena such as seismic activity. In this context, we apply our new methodology to real earthquake data, showing an efficient estimation of the model parameters, along with improvements in both accuracy and computational speed compared to existing methods.
Keywords: Point process modeling; Spatio-temporal Hawkes processes; Bayesian EM; Inlabru