12 noon - 1pm
Thursday 7 February 2019
Predictive policing: myth or reality?
Speaker: Sylvain Delahaies (Surrey)
Predictive policing aims at forecasting where and when crime will take
place in the future. While millions are currently invested in unproven
predictive softwares we focus on self-excited point processes which have
recently become popular to predict crime. We consider a police patrol
allocation experiment and investigate under what parameter regimes the
model might prove useful. Then we consider a novel Bayesian sequential
data assimilation algorithm for joint state-parameter estimation by
deriving an approximation Poisson Gamma Kalman filter. Finally we apply
the data assimilation scheme and police patrol experiment to real
Chicago crime data.
This event has passed