Self-exciting point process modelling of crimes on linear networks

Although there are recent developments for the analysis of first and second-order characteristics point processes on networks, very few attempts in introducing models network data. Motivated by crime data Bucaramanga (Colombia), we propose a spatiotemporal Hawkes process model adapted to events living linear networks. We consider non-parametric modelling strategy, which follow estimation both background triggering components. Then semi-parametric version, including parametric based covariates, one effects. Our can be easily multi-type processes. outperforms planar improving fitting self-exciting model.