Metro passenger flow prediction: a double-stage decomposition combined with Enhanced-BiGRU model considering multiple factors

Accurate and reliable passenger flow prediction is essential to formulate operational plans, improve service quality, and alleviate traffic pressure for the metro transportation system. However, passenger flow sequences exhibit strong nonlinearity and random fluctuations, posing significant challenges to prediction. To address this problem, we propose a metro passenger flow prediction model based on double-stage decomposition and Enhanced-BiGRU considering multiple factors. Firstly, the double-stage decomposition module is constructed to reduce the volatility of passenger flow sequences. Specifically, the first-stage decomposition decomposes and reconstructs the original sequence into high-frequency, mid-frequency, and low-frequency components. The second-stage decomposition further reduces the sequence volatility by decomposing complex high-frequency component. Furthermore, an influence factor selection module is introduced to select the important factors that affect passenger flow. Finally, we design a passenger flow prediction module called Enhanced-BiGRU, in which convolutional neural network (CNN) and attention mechanism (AM) are employed to overcome the deficiency in the feature extraction capacity of BiGRU, and sparrow search algorithm (SSA) is leveraged to optimize the hyperparameters of BiGRU. Experimentally, the proposed model is validated on the Hangzhou Metro dataset and achieves superior performance compared to the state-of-the-art.