Traffic congestion poses significant socio-economic challenges, and many urban commuters experience time and cost burdens due to traffic jams. Effectively managing and predicting traffic congestion is crucial for contemporary urban planning and operations. This study aims to predict congestion using time-series deep learning models, focusing on a chronically congested road section. The research area is the Atlanta Downtown Connector, a heavily trafficked route passing through the heart of the city of Atlanta. To address issues often associated with sensor and GPS-based data collection, traffic data was obtained from an open-source online map service, Google Maps. A multivariate time-lagged LSTM model was employed for traffic congestion prediction, taking into account the seasonality of traffic data. The results of predicted traffic congestion patterns revealed similarities to historical data, with the prediction model achieving low error rates. This research has the potential to assist policymakers involved in traffic management and provide valuable information for citizens planning their journey on urban areas.
"Prediction of Traffic Congestion Using a Time-Series Model and Spatiotemporal Data: A Case Study of the Atlanta Downtown Connector,"
The Geographical Bulletin: Vol. 64:
2, Article 7.
Available at: https://digitalcommons.kennesaw.edu/thegeographicalbulletin/vol64/iss2/7