Short-Term Passenger Flow Prediction of Railway Station Based on Improved LSTM

Accurately predicting passenger arrivals helps ease the pressure on rail passenger station operations. In order to realize the traffic forecasting of railway passenger station, according to the time distribution characteristics of passenger flow, traffic forecasting method of railway passenger station is studied. Using the ticket data as the research object, this paper research the time distribution of passenger traffic flow in railway station under different time dimensions, hierarchical clustering algorithm and threshold clustering algorithm are taken to cluster the passenger flow, and the improved LSTM traffic forecasting model was built. Aiming at LSTM model input data hierarchical segmentaion scale is large, which lead to the depth of network layers is not enough, we rebuild the LSTM model. The method is verified by the actual AFC data of QingHe Station, the prediction results of traditional prediction model are compared. It shows that: the improved model has the higher prediction accuracy than other traditional forecasting models, mean absolute error is 10% lower. It shows that this method can better describe the time correlation of passenger flow, and deeply mine the internal mechanism of passenger flow change.