Short-term Passenger Flow Prediction of Metro based on ARIMA and LSTM Models

The essay studies the comparative effectiveness of ARIMA and LSTM models in predicting passenger loads in metro systems, emphasizing the Beijing metro as a case study. It has been argued that accuracy is essential when it comes to forecasting passenger flows, which will, in turn, enhance the efficiency of urban mass transit systems and promote a comfortable and preferable traveling experience for the passengers. The article meticulously dissects and compares the techniques proposed in this paper in a detailed fashion by revealing each of their strengths and weaknesses and historical passenger load numbers. The techniques are illustrated using data from the AFC system of the Beijing Metro, focusing particularly on Wangfujing Station for its tremendously high passenger flow. The effectiveness of these techniques is shown on a variety of evaluation metrics, including mean squared error (MSE), mean absolute error (MAE), root mean squared error (RMSE), and R². It is found that the ARIMA model is better suited to capture shorter intervals (2 minutes) because of its lower error rates and that it leads to greater clarity in longer intervals (5 minutes) despite having higher error rates. This research contributes to the broader field of intelligent transportation systems by providing insights into the effective application of predictive modeling techniques for enhancing metro system operations.