Crash prediction model based on Pi-sigma fuzzy neural network

A crash prediction model based on the Pi-Sigma fuzzy neural network was put forward. The Takagi-Sugeno inference system and BP neural network were employed in the model. AADT, traffic load, design speed and lane width were taken as input variables while the number of crashes per kilometer per year as the output variable. The model was further calibrated and validated with the data which includes 133 segments from arterial road in Harbin City and 5 years' crashes. Then, the model was compared with the fuzzy logic model and the BP neural network model. The results show that the Pi-Sigma fuzzy neural network crash prediction model is better than the other two models from the aspects of prediction accuracy and calculation efficiency, which makes it more suitable for the rapid crash prediction with big sample data.