Abstract: Driver's factors are very important for the traffic safety. Questionnaire and BP neural network are used, the neural network structure with different hidden layer, neuron number and transfer function is established considering many parameters, such as age, gender, accumulative driving time, physiological conditions, etc. The optimal neural network structure is obtained to predict the traffic accident rate. The results show that the neural network is available for the prediction of accident rate and the driver subgroups with accident proneness are identified.