Aiming at the difficulty in predicting the duration of highway traffic accidents, a traffic accident duration prediction model based on factor analysis and back propagation neural network(BPNN) is established. Based on real traffic accident data collected from Baomao Expressway, factor analysis is utilized to extract the common factors influencing the duration of traffic accidents. The extracted public factors are taken as the input, and BP neural network is used to predict the duration of traffic accidents. The experimental results show that compared with the typical regression algorithm and the support vector machine, the improved BPNN method with factor analysis proposed in this paper can not only improve the prediction accuracy by7.8%, but also solve the problems of low data processing efficiency and slow iteration speed of traditional BP neural network.