In cities, traffic congestion is a major problem and traffic forecast is critical for effective incident response. Machine learning and deep learning approaches have been used to anticipate traffic flow, speed, and occupancy, but there are obstacles due to traffic flow's time and space interdependence. This research provides an advanced deep learning framework for forecasting traffic flow, speed, and occupancy on a heavily traveled highway throughout the year. Based on incoming and historical data streams, the system incorporates an outlier detection and anomaly compensation approach. Deep learning architectures such as convolutional neural networks (CNNs), recurrent neural networks (RNNs), and hybrid spatiotemporal modeling (CNN-LSTM) use spatial and temporal characteristics acquired from 36.34 million data points. The models consistently outperform traditional methods, and a comparative analysis is undertaken to determine the appropriate time horizon of historical data required to predict traffic flow at various future time periods. Based on recent observations and historical data, the results reveal that deep learning can provide real-time traffic forecast, future traffic prediction, and short-term traffic prediction.
Traffic Congestions Prediction Using Machine Learning and Deep Learning Techniques
2023-12-12
459220 byte
Conference paper
Electronic Resource
English
A Deep Learning Approach to Analyze Traffic Congestions for Effective Traffic Management
Springer Verlag | 2022
|Airports facing air traffic congestions
SLUB | 1989
|Spatio-Temporal Evolution of Traffic Congestions on Urban Freeways
British Library Conference Proceedings | 2009
|