In considering major contributions to traffic accidents, it has been pointed out that they may be accounted for by drivers' errors in their information-processing characteristics; cognition and decision making. Knowing the information mechanism of drivers' errors may provide important aspects of developing and evaluating safe driving systems.

We address the prediction of drivers' errors using computer simulation program and propose a neural network based driver model for the purpose of predicting drivers' errors. From the experiments using a driving simulator, the subjects' behavior was modeled by the neural networks from operational data in time series. Also, the driver model made it possible to predict the timing delay in braking operation of the subject, that is drivers' error, when the preceding car had no stopping lamps in a stopping situation. Hence, we presented the feasibility of the neural network based driver model. Moreover, by analyzing the input-output relationship of the neural networks, we found that the signals of the tail-lamps and the size of the preceding car were available information when deciding the braking operation.