Short-time traffic flow prediction based on HSFLSTM integrating EMD and Mogrifier

In transportation management, short-term traffic flow prediction plays a crucial role in road planning and system layout. This study proposes a hybrid Hidden State Fusion Long Short-Term Memory (HSFLSTM) model that integrates Empirical Mode Decomposition (EMD) and Mogrifier to predict short-time traffic flow. By decomposing traffic data using EMD, relatively stable sub-sequences are obtained. The Mogrifier is then employed to interact with input and hidden states, enhancing their correlation. The HSFLSTM model combines these two methods for training and predicting short-term traffic flow data. Experimental results demonstrate that on the PeMS dataset, the model achieves a MAPE evaluation metric of 4.839%, significantly outperforming other models.