Network-Wide Evacuation Traffic Prediction in a Rapidly Intensifying Hurricane from Traffic Detectors and Facebook Movement Data: Deep-Learning Approach

Abstract

Traffic prediction during hurricane evacuation is essential for optimizing the use of transportation infrastructures. Traffic prediction can reduce evacuation time by providing information in advance on future congestion. However, evacuation traffic prediction can be challenging because evacuation traffic patterns are significantly different than are those for regular period traffic. A data-driven traffic prediction model is developed in this study by utilizing traffic detector and Facebook movement data during Hurricane Ian, a rapidly intensifying hurricane. We select 766 traffic detectors from Florida’s four major interstates to collect traffic features. Additionally, we use Facebook movement data collected during Hurricane Ian’s evacuation period. The deep-learning model is first trained on regular period (May to August 2022) data to understand regular traffic patterns. Then, Hurricane Ian’s evacuation period data are used as test data. The model achieves 95% accuracy (RMSE=356) during regular period but underperforms with 55% accuracy (RMSE=1,084) during the evacuation period. Then, a transfer learning approach is adopted in which a pretrained model is used with additional evacuation-related features to predict evacuation period traffic. After transfer learning, the model achieves 89% accuracy (RMSE=514). Adding Facebook movement data further reduces the model’s RMSE value to 393 and increases accuracy to 93%. The proposed model is capable of forecasting traffic up to 6-h in advance. Evacuation traffic management officials can use the developed traffic prediction model to anticipate future traffic congestion in advance and take proactive measures to reduce delays during evacuation. Hurricane evacuation causes significant traffic congestion in transportation networks. Increased traffic demand can affect the evacuation process because it delays the movement of people to safer locations. To remedy this issue, an accurate traffic prediction model is beneficial for evacuation traffic management. The prediction model can give expected traffic volume on evacuation routes well in advance, which allows traffic management agencies to prepare for and activate strategies such as emergency shoulder utilization, adjustments to signal timing for optimal traffic flow, and others on those evacuation routes. This work aims to construct a data-driven model to predict traffic flow with a lead time of up to 6 h. The model can be used to forecast networkwide traffic in real time. Thus, practitioners can use this tool to effectively implement evacuation traffic management strategies by determining the timing, locations, and extent of those strategies based on predicted traffic volume. Another benefit of this model is that it can be trained with data from normal period and historical hurricane evacuations and then be implemented for future hurricanes.