Traffic Flow Prediction Using Signal Decomposition and Deep Learning Techniques

Accurate prediction of traffic flow is essential for urban management, not only to alleviate congestion, but also to optimize resource allocation and enhance citizens' travel experience. However, traditional models often face difficulties in dealing with nonlinear and non-stationary traffic data, especially in complex time series prediction, and the accuracy is low. In this paper, a model combining signal decomposition, convolutional neural network (CNN) and bidirectional long Short-Term memory network (BiLSTM) is proposed, which can better capture the multi-frequency characteristics and temporal dependencies in traffic data. Experimental results show that the proposed model performs well in 6-step, 12-step and 24-step traffic flow prediction, and the error index is significantly lower than other models, showing the efficiency and stability in short-term and long-term traffic prediction.