Abstract: By analyzing the driving characteristics of public transport vehicles and driving environment,the following key factors that affects bus travel time are determined: the weather,the time period (the peak / flat peak), traffic flow and the length of the road.Combined with nonlinear BP neural network’s advantages such as mapping ability, self-learning and adaptive ability and generalization ability and firefly algorithm’s advantages such as less parameters,simple operation and good stability,a BP neural network optimization method is put forward that is optimized by the firefly algorithm to reduce the training time of neural network and can improve the stability of prediction. Then, GPS datas and sampling datas are used to build a model to realize the accurate prediction of travel time of transit vehicles.MATLAB software is used for simulations.The optimized algorithm,the classical BP neural network algorithm and Kalman algorithm are compared in the simulations which show that the optimized prediction model has higher accuracy and is more stable for bus travel time prediction.

Key words: intelligent transportation system, public traffic, firefly algorithm, BP neural network, Kalman algorithm, travel time prediction