A FUZZY NEURAL NETWORK MODEL OF JUDGE DRIVER PRONENESS

A comprehensive evaluation expert system of driver proneness is established in this paper. The system includes a Bp network, a fuzzy inferring engine and a knowledge library. Eight index values (which are high speed judgement, speed tracing, dark adaptation, tracking, selective response, attention, kinesthetic vision and depth perception) are used as characteristic parameters to form the testing samples. Using K-means method cluster the testing samples beforehand, standard learning samples are formed and then are used to train and adjust the system. The experiment proves that the evaluation effect of driver proneness is satisfying when we use the trained system to evaluate the driver proneness within driver's mental and physical parameters.