LEAM: A Large-scale Events Aware Module for Multi-step Intercity Flow Prediction

The accurate multi-step prediction of passenger flow is crucial for the efficient operation of intercity Intelligent Transportation Systems (ITS). However, the presence of large-scale anomalous events can significantly impact the accuracy of these predictions, particularly in the case of intercity events. In this paper, we propose a novel approach that utilizes a spatio-temporal graph to predict passenger flow. Our method incorporates population migration data, allowing us to analyze the influence of large-scale anomalous events on intercity public transportation. Additionally, we introduce a large-scale events aware module (LEAM) designed to detect and evaluate the impact of anomalous events on passenger flow. Our analysis provided evidence to support the rationality of the proposed architecture. To evaluate the performance of our approach, we employ three popular deep learning models for multi-step prediction. The experimental results demonstrate that our architecture significantly improves the accuracy of anomalous event prediction and enhances the global optimization of predictions, compared to the same models without the integration of LEAM.