Forecasting grey model of traffic accidents based on variable parameter value rolling

Taking the death rates from 1990 to 2006 in Chinese accidents as basic datum, a variable parameter value rolling grey forecasting model (VRGM) is proposed based on the fixed parameter value rolling grey forecasting model (FRGM), which has a constant parameter and equals 0.5. It is apparent that if the parameter value is decided in consideration of the factors, such as the number of population, the number of vehicles and the length of highway, which influence the forecasting objective, better result will be achieved. According to death rates per 10 thousand vehicles in China, a VRGM (1, 1) model is constructed. The average residual error of FRGM (1, 1) model, VRGM (1, 1) model and VRGM (1, 1) model plus 3-points average are 6.97%, 5.67% and 4.61% respectively. The research suggests that VRGM (1, 1) model is better than FRGM (1, 1) model, and VRGM (1, 1) model plus 3-points average can further smooth the prediction results and increase the reliability. 3 tabs, 15 refs.