A hybrid neural network for large-scale expressway network OD prediction based on toll data

Area of Science:

• Traffic Engineering
• Artificial Intelligence
• Data Science

Background:

• Accurate Origin-Destination (OD) prediction is crucial for effective traffic monitoring and management.
• Expressway toll systems generate extensive trip data suitable for OD prediction.
• Existing models often lack the accuracy and flexibility needed for complex urban traffic networks.

Purpose of the Study:

• To develop a novel neural network for accurate OD prediction using expressway toll data.
• To enhance prediction accuracy by incorporating city attributes like GDP and population.
• To improve model applicability by categorizing cities based on economic and demographic scales.

Main Methods:

• Development of the Expressway OD Prediction Neural Network (EODPNN) with Feature Extension, Memory, and Prediction modules.
• Integration of city attributes (GDP, population, vehicle numbers) into the neural network.
• Classification of cities into categories based on economy and population for tailored predictions.

Main Results:

• The EODPNN demonstrated superior prediction performance compared to traditional models (ARIMA, SVM) and Bidirectional Long Short-term Memory networks.
• Incorporating city attributes significantly increased the accuracy of OD predictions.
• City categorization enhanced the model's applicability and predictive power.

Conclusions:

• The proposed EODPNN offers a significant advancement in OD prediction accuracy and flexibility.
• The method is verified and shows potential for application in other OD data-based traffic management systems.
• This approach provides a more accurate and adaptable solution for traffic planning and operational decisions.