Assault crime dynamic chain event graphs

In this paper a new class of graphical Bayesian discrete semi-Markov models is developed to describe the various pathways that might lead someone into perpetrating various kinds of crime involving assault and violence. Our discrete probability models are crafted to embody various theory and empirical studies by psychologists and sociologists explaining and describing this development. This probability model is then used to formally structure a new decision support system to help authorities evaluate public risks guided by not only archived but real time data. We argue here that such systems will be able to provide provisional quanti ed evaluations of the impacts of various policy and policing decisions into the short and medium term. The construction of such probability models is illustrated throughout by examples. We end the paper with a more detailed description of a model built to support authorities to frustrate populations of criminals who have been radicalised into violent extremism