Stochastic-physics-of-crime analyses for TPS data
Source:R/tps_stochastic.R
morie_tps_stochastic.RdR port of morie.tps_stochastic. Four jurisdiction-agnostic
callables: temporal-only exponential Hawkes self-exciting fit,
seasonal ARIMA forecast on monthly counts, Euler-Maruyama
Ornstein-Uhlenbeck simulation, and a 1-D Fokker-Planck density
evolution. The R port keeps optimisation in base R
(stats::optim) and the seasonal forecast in
stats::arima so no external time-series package is needed.
Details
All functions return a multi-section morie_rich_result list.
Functions
morie_tps_hawkes_temporal_fit: fit mu, kappa, omega of an exponential-kernel Hawkes process to incident times; report branching ratio + AIC/BIC.morie_tps_sarima_forecast: seasonal ARIMA on monthly counts with train / hold-out MAPE.morie_tps_langevin_simulate: Euler-Maruyama OU SDE paths fitted to daily counts.morie_tps_fokker_planck_grid: 1-D finite-difference density evolution under OU drift+diffusion.
References: Mohler et al. 2011 (self-exciting point process crime); Short, D'Orsogna, Bertozzi 2010 (stochastic physics of crime).