In real-time traffic modeling, selecting the appropriate update window is crucial. Updating too frequently can lead to unnecessary computational burden, while not updating enough can result in outdated information. We propose a novel framework called Outlier Weightage Autoencoder Modeling (OWAM), which efficiently manages larger datasets and more complex scenarios with less computational load than previously used techniques. OWAM incorporates the Long Short-Term Memory (LSTM) model, widely used in traffic prediction, while adapting to changing traffic patterns by analyzing data before selecting a specific update window.
Key Components
- Flow Probability Distributions (FPDs): FPDs represent the probability of traffic occurrences within specific time intervals derived from real traffic data. They encapsulate traffic distribution patterns and can be created for vehicle flow, average speed, or any relevant traffic data to gain a comprehensive understanding of traffic dynamics.
- Outlier Weightage Autoencoder Modeling (OWAM): OWAM is a novel framework that combines the LSTM model with outlier weightage autoencoders to improve performance and efficiency in real-time traffic modeling. It adapts to changing traffic patterns by analyzing data before selecting a specific update window, ensuring optimal performance.
Benefits
- Improved Performance: OWAM shows superior performance compared to previous outlier-based and baseline techniques, with faster processing times than GNN-based techniques like DGCRN, making it the preferred choice for real-world applications.
- Scalability: OWAM’s ability to efficiently manage larger datasets and more complex scenarios with less computational burden than DGCRN positions it as a scalable solution for traffic management and similar real-time applications.
Conclusion: In conclusion, selecting the appropriate update window is critical in real-time traffic modeling. Our proposed framework, OWAM, incorporates the LSTM model and adapts to changing traffic patterns by analyzing data before selecting an optimal update window, resulting in improved performance and scalability compared to existing techniques. By leveraging FPDs and outlier weightage autoencoders, OWAM enables efficient management of larger datasets and more complex scenarios, making it a valuable tool for real-world applications.
