An urban road traffic flow prediction method based on multi-information fusion

Area of Science:

• Transportation Science
• Artificial Intelligence
• Data Science

Background:

• Accurate traffic flow prediction is crucial for intelligent transportation systems.
• Existing methods often focus on short-term predictions and struggle to incorporate diverse external factors.
• Spatio-Temporal Graph Neural Networks (STGNN) capture spatial and temporal dependencies but overlook long-term trends and cyclical patterns.

Purpose of the Study:

• To propose a novel Multi-information Fusion Prediction Network (MIFPN) for enhanced traffic flow forecasting.
• To effectively integrate historical traffic data, external factors, and both long-term and short-term temporal features.
• To improve the accuracy and reliability of traffic flow predictions, especially for longer time horizons.

Main Methods:

• Utilized a subsequence converter to learn temporal relationships from extended historical sequences incorporating external information.
• Employed a superimposed one-dimensional inflated convolutional layer for long-term trend extraction.
• Implemented a dynamic graph convolutional layer for periodic feature extraction and a short-term trend extractor for immediate temporal dynamics.
• Fused extracted long-term trends, cyclical features, and short-term features for final predictions.

Main Results:

• The MIFPN model demonstrated significant improvements in traffic flow prediction accuracy.
• Achieved an average improvement of 11.2% over baseline models in long-term predictions (up to 60 minutes).
• Successfully integrated diverse data sources, including historical traffic data and external factors like weather and POI distribution.

Conclusions:

• The proposed MIFPN effectively captures complex temporal dependencies and fuses multi-source information for superior traffic flow prediction.
• The model's ability to extract both long-term trends and short-term features, alongside cyclical patterns, leads to enhanced forecasting accuracy.
• MIFPN offers a promising advancement for intelligent transportation systems requiring reliable long-term traffic flow predictions.