A neural network for traffic flow prediction with parallel processing of expanded convolutional and radial networks

Area of Science:

• Artificial Intelligence
• Machine Learning
• Data Science

Background:

• Traffic flow prediction is complex due to numerous influencing factors.
• Traditional time series models often lack accuracy.
• Temporal Convolutional Networks (TCNs) offer parallel processing but struggle with raw data filtering and real-time capabilities.

Purpose of the Study:

• To propose a novel model, Radial Temporal Convolutional Networks (RSCN), for enhanced traffic flow prediction.
• To address the limitations of existing TCNs in handling raw data and real-time processing.

Main Methods:

• Developed RSCN by integrating a radial network and inflated convolution into the TCN architecture.
• Employed a multi-layer processing mechanism with residual layers and dimensionality reduction for feature extraction.
• Utilized parallel processing for efficient data handling.

Main Results:

• RSCN demonstrated superior efficiency compared to several existing traffic flow prediction methods.
• The proposed model effectively extracts features from raw traffic data, overcoming TCN limitations.

Conclusions:

• RSCN offers a significant advancement in traffic flow prediction accuracy and efficiency.
• The model's architecture is well-suited for handling the complexities of real-time traffic data analysis.