Traffic-Data Recovery Using Geometric-Algebra-Based Generative Adversarial Network.

Area of Science:

• Artificial Intelligence
• Data Science
• Transportation Engineering

Background:

• Accurate traffic data is crucial for traffic prediction, congestion analysis, and road network planning.
• Existing methods struggle with significant traffic data loss.
• Geometric algebra offers powerful capabilities for high-dimensional data processing.

Purpose of the Study:

• To propose a novel geometric-algebra-based generative adversarial network (GAN) for repairing missing traffic data.
• To leverage the representational power of geometric algebra for multidimensional traffic parameter correlation learning.
• To enhance the accuracy and robustness of traffic data recovery.

Main Methods:

• Developed a GAN with a generator comprising geometric algebra convolution, attention, and deconvolution modules.
• Implemented a loss function incorporating global and local data mean squared errors.
• Utilized a multichannel convolutional neural network for the discriminator to optimize adversarial training.

Main Results:

• The proposed geometric-algebra-based GAN effectively repaired missing traffic data using real-world highway datasets.
• The method demonstrated robust performance in traffic data recovery.
• Experimental results showed superior performance compared to state-of-the-art methods.

Conclusions:

• The geometric-algebra-based GAN is a highly effective approach for traffic data recovery.
• This method offers a robust solution for handling large amounts of missing traffic data.
• The findings suggest significant improvements in traffic data imputation accuracy and reliability.