The recognition and prediction of intake air flow was built based on wavelet networks due to the nonlinear and dynamic property of engine intake system. To improve the reliability and precision of wavelet network model, the parameters and control law were learned and optimized with Davidon least square (DLS) algorithm. Then BP neural network model for intake air flow under transient conditions was established and compared with wavelet network model based on the actual acquisition data. The results show that the wavelet network model can successfully forecast intake air flow of gasoline engine under transient conditions and is superior to BP neural network model due to higher accuracy. Accordingly, the model may apply to the accurate control of transient air fuel ratio.