Accurate and stable prediction of traffic flow in a short period of time is an important link in the realization of intelligent traffic control system, which is of great significance for real-time adjustment of intersection signal control scheme. In view of this, a prediction method of WD-BNN joint model is proposed, and MAPE and RMSE are introduced as evaluation indexes to evaluate the model from two aspects of accuracy and stability. The results show that MAPE and RMSE of WD-BNN combined model are smaller than Wavelet network, Bayesian network, l-m network and other methods under different time prediction scales of 5min, 10min, 15min, etc., and the accuracy and stability of results have been improved to varying degrees.