Short-Term Traffic State Prediction Based on the Critical Road Selection Optimization in Transportation Networks.

Area of Science:

• Transportation Engineering
• Artificial Intelligence
• Network Science

Background:

• Short-term traffic prediction is crucial for managing large-scale transportation networks.
• Corrupted or missing data presents significant challenges in traffic state prediction.
• Critical roads significantly influence traffic conditions on adjacent routes.

Purpose of the Study:

• To propose a novel hybrid method for short-term traffic state prediction in large-scale networks.
• To optimize the selection of critical roads for enhanced prediction accuracy.
• To address challenges posed by corrupted or missing traffic data.

Main Methods:

• Developed a utility function for critical road Quality of Service (QoS) based on coverage and data score.
• Formulated a critical road selection optimization model to maximize QoS utility within resource constraints.
• Introduced an innovative critical road selection approach considering network topology and urban mobility.
• Utilized a convolutional long short-term memory (ConvLSTM) neural network with critical road traffic speed as input.

Main Results:

• The proposed method demonstrated superior performance compared to existing deep learning (DL) approaches.
• Experiments on the Beijing traffic network validated the effectiveness of the critical road selection strategy.
• The method shows improved traffic prediction accuracy, especially when critical road sections are considered.

Conclusions:

• The hybrid traffic prediction method effectively leverages critical road selection optimization.
• The approach enhances the accuracy of short-term traffic state prediction in complex urban networks.
• This study offers a robust solution for traffic prediction challenges with incomplete data.