Skip to Main Content
In view of the disadvantages of GM(1, 1) model, which has low precision for prediction of violent non-stationary data series. Based on availability and characteristics of accident data, the GM(1, 1) improved model and Verhulst improved model are put forward by using metabolism methods. Application of traffic accident data from 1996~2009 of Heilongjiang Province shows that the Verhulst improved model is more precise than GM(1, 1) improved model for violent changing data series. The average precision of Verhulst improved model can reach 85%, which can basically meet the demand for road traffic safety management.