An Information Renewal GNN Model for Road Traffic Accident Forecasting

Abstract

Road traffic system is a complicated non-linear system, in which road traffic accident is considered as the behavioral characteristic variable, whose developmental changes have trends of increase and stronger random fluctuation. Considering this point, the authors establish a combination forecasting model (namely GNN) based on grey forecasting model and ANN. First, grey information renewal GM (1,1) is established based on GM (1,1) and used to forecast the change tendency, then BP network is applied to modify grey residuals to capture stochastic phenomenon. The results show that the dual character of the road traffic time series with trends of increase and random fluctuation can be better described. Meanwhile, with advantages of GM (1,1) and ANN, GNN has obtained better forecasting precision than single grey information renewal GM (1,1). In summary, GNN can be applied as a novel, practical and simple forecasting tool in road traffic accident forecasting.