Road Congestion Prediction Using Adaptive Graph Expert Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing congestion prediction models using deep learning suffer from poor interpretability, limited ability to handle highly dynamic and heterogeneous urban traffic data, and lack robustness against data noise.
Innovation Solution
A method and apparatus for constructing a road congestion prediction model that defines a road traffic network as a directed weighted graph, utilizes a mixture of adaptive graph learners (MAGL) model, trend expert model, and periodic expert model to enhance interpretability and robustness, incorporating static road attributes and real-time vehicle track data for accurate prediction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If deep learning modeling is used for congestion prediction, then prediction accuracy can be improved, but model interpretability deteriorates
Solution Approach 1:
The model is segmented into multiple specialized expert models (trend expert, periodic expert, spatial-temporal expert) that each handle specific aspects of traffic prediction. This segmentation allows the system to maintain high prediction accuracy through specialized components while improving interpretability by making each expert's function and decision logic transparent and analyzable independently.
Solution Approach 2:
A gating mechanism serves as an intermediary that routes input data to appropriate expert models based on traffic conditions. This mediator layer provides interpretability by showing which experts are activated and why, while maintaining prediction accuracy by directing data to the most suitable expert for each situation.
2Measurement precision
If deep learning modeling with high spatiotemporal dependencies is used, then prediction capability can be improved, but model complexity increases
Solution Approach 1:
The complex prediction task is divided into separate expert models handling different spatiotemporal aspects (trend, periodicity, spatial-temporal patterns). This segmentation reduces overall model complexity by allowing each expert to be simpler and more specialized, while maintaining high prediction capability through their coordinated work.
Solution Approach 2:
The model dynamically selects and activates different expert models based on current traffic conditions through the gating mechanism. This dynamic approach allows the system to handle complex spatiotemporal dependencies only when necessary, reducing average model complexity while maintaining high prediction capability when complex patterns are present.
3Productivity
If existing congestion prediction models are used, then prediction function can be provided, but robustness against data noise deteriorates
Solution Approach 1:
By segmenting the prediction function into multiple expert models with different specialization (trend, periodic, spatial-temporal), the system gains robustness against noise because each expert can be optimized to handle specific types of data patterns and noise characteristics, rather than trying to handle all patterns in a single vulnerable model.
Solution Approach 2:
The gating mechanism provides feedback by routing data to experts based on learned patterns and can adjust which experts are activated based on data quality and noise levels. This feedback mechanism enhances robustness by allowing the system to rely more on noise-resistant experts when data quality is poor while maintaining prediction functionality.
Data Source
AI summary
Provided are a method and an apparatus for constructing a road congestion prediction model, a device, a medium, and a product. A road traffic network is defined as a directed weighted graph. Historical dynamic traffic features of each road segment in the road traffic network are obtained as sample data, including recent dynamic traffic features and periodic dynamic traffic features. The sample data is input into a mixture of adaptive graph learners (MAGL) model for learning, and a probability prediction vector is output. The sample data is input into a trend expert model, and a trend distribution vector of a predicted probability of future traffic conditions is output. The periodic dynamic traffic features are fused to determine a periodicity prediction vector. An aggregated logit vector is obtained. An objective function is determined based on the aggregated logit vector. Congestion prediction training is performed to obtain a road congestion prediction model.