Prediction of short-term passenger flow of subway based on LSTM model

Abstract

With the continuous development of urban rail transit, the subway is becoming the travel choice for more and more people. And the increase in travel demand also brings about the problem of subway congestion. This paper will use the subway card swipe data of Hangzhou City from January 1st to 20th, 2019, which were counted after data cleaning. With root mean square error (RMSE), mean absolute percentage error (MAPE), coefficient of determination (R^2), and loss values as standards, the long short-term memorys (LSTM) parameters were adjusted. The LSTM model is trained with time and passenger flow data, and the model is used to predict the short-term passenger flow in and out of subway stations. From the prediction results, the LSTM model has a good adaptability to the subway station short-term passenger flow. The trend of the predicted line is the same as the real value, and the difference is also controllable. The short-term passenger flow prediction of subway stations is beneficial to reduce the loss caused by congestion and improve the travel efficiency of residents.