Optimized GM(0,N) model with exponential-trigonometric transformations and PSO for queue length prediction at metered...

Area of Science:

• Traffic Engineering
• Transportation Science
• Data Modeling

Background:

• Conventional Grey Models (GM) struggle with time series prediction accuracy due to unconsidered influencing factors.
• Predicting queue lengths at metered roundabouts is crucial for traffic management.

Purpose of the Study:

• To develop an optimized GM(0,N) model for enhanced queue length prediction at metered roundabouts.
• To improve prediction accuracy by integrating traffic entry volume, conflicting flow, and signal timing.

Main Methods:

• An optimized GM(0,N) model was developed, incorporating exponential and trigonometric sequence transformations.
• Particle Swarm Optimization (PSO) was used to determine optimal model parameters.
• The model was validated using real-world data from metered roundabouts in Adelaide, Australia.

Main Results:

• The proposed optimized GM(0,N) model demonstrated superior prediction accuracy compared to An's model, GM(1,1), GM(1,N), and conventional GM(0,N) models.
• Accuracy was assessed using Mean Relative Error (MRE), Root Mean Square Error (RMSE), and Mean Absolute Error (MAE).
• Box plot analyses confirmed the enhanced model's superior performance.

Conclusions:

• The optimized GM(0,N) model offers a more accurate approach to predicting queue lengths at metered roundabouts.
• The model is effective for managing unbalanced roundabout traffic and guiding detector placement.
• Enhanced sequence transformation and parameter optimization are key to improving GM model performance in traffic prediction.