Macro Prediction Model of Road Traffic Accident Based on Neural Network and Genetic Algorithm

Qin Liyan,Shao Chunfu

DOI: https://doi.org/10.1109/icicta.2009.93

2009-01-01

Abstract:The whole society has paid more and more attention to the road traffic safety in China. In order to scientifically and accurately predict the future road traffic safety situation, evaluation indexes and main influence factors of road traffic safety were analyzed. Then traffic accident death toll was selected as the evolution index and three factors were taken as parameters, including motor vehicle amount, highway mileage and per capital gdp. The two prediction models of BP neural network and hybrid algorithm based on neural network and genetic algorithm were built respectively. The neural models used the traffic accident data from 1978 to 1998 as training samples. And the data from 1999 to 2004 were used to check the models. Forecasting result shows that neural network model based on genetic algorithms has higher prediction accuracy and more network generalization than those of BP neural network; the forecasting values of accident death toll in 2010 and 2020 are 139 thousand people and 167 thousand people respectively.