An improved particle swarm optimization algorithm based urban rail passenger flow prediction model: a case study in Beijing, China

To accurately predict the passenger flow of urban rail transit under different external conditions, the distribution characteristics of urban rail passenger flow were analyzed using the AFC data of rail transit, and three influencing factors of urban rail passenger flow were extracted. The typical support vector regression (SVR) algorithm was utilized to construct the passenger flow prediction model of the urban rail, then we proposed an improved particle swarm optimization (IPSO) algorithm for optimizing the prediction model. Finally, the prediction accuracy of the proposed model was verified by comparative analysis. The results show that the IPSO-SVR passenger flow prediction model has better prediction accuracy compared with the SVR and long short-term memory networks (LSTM) models, also the traditional grid optimization method, the mean square error (MSE), and relative accuracy (RA) are 1.54% and 92.37%. The validation of model performance is carried out compared with three other models in a previous study. There is a positive correlation between the model prediction errors and the scale of urban rail transit passenger flow. The three influencing variables of the time segments, working day type, and weather can also effectively characterize the coupling characteristics of urban rail passenger flow, and improve the model prediction accuracy. The results have important application value for evaluating passenger flow status, improving the operation quality, and operation organization of urban rail transit.