A serious crime flash-up forecast based on hierarchy of criminality behavior

This work is devoted to a quest of criminality forecast methods based on universal behavior rules of complex nonlinear hierarchical systems which are poorly amenable to routine modelling. Using a weekly criminality data from Yaroslavl for 1993–2001 years we study crimes’ dynamics and propose serious crime flash-up forecast algorithms forecasting about 70% of flash-ups during 30% of alarm time with 4% of false alarms.