Abstract
In intelligent transportation systems, real-time and accurate short-time traffic flow prediction is the focus of research. Real-time and accurate short-time traffic flow prediction is the basis of intelligent traffic control and management. However, due to the high complexity, randomness, and uncertainty of short-time traffic flow, the effect of prediction is not satisfactory. To cope with the complexity, randomness and uncertainty of short-term traffic flow prediction, we propose a prediction model to achieve the desired accuracy. The wavelet neural network (WMN) is developed to denoise and reconstruct the data, and WNN model is used to predict short-term traffic flow. To avoid the WNN model from falling into the local optimal situation during training, an ant colony algorithm is combined to find the optimal values for the weights. Experimental results show that the combined WNN model can achieve a high accurate prediction for short-time traffic data prediction.