Abstract: The mining of floating car data is one of the research methods widely used in the transportation field. The basic BP neural network is also used for the prediction of traffic flow. In this paper, wavelet transform is used to decompose and reconstruct the low-frequency signal and high-frequency signal. Combining improved Firefly Algorithm to optimize basic BP neural network, a prediction model of speed of road network traffic flow is constructed. The model is trained by using the floating car data of urban road network and the model prediction results are empirically analyzed by the test data. It is proved that the accuracy of the model for the traffic velocity prediction of the road network at a specific moment is 46.56% higher than that of the BP neural network algorithm and the stability of the forecast of traffic flow speed within 24 hours of the road network increases by 39.08%.

Key words: floating car data, wavelet transform, Firefly Algorithm, BP neural network, speed prediction