Intelligent trapezoid and variable weight combination-based reconstructed GM model.

Area of Science:

• Mathematical modeling
• Time series analysis
• Predictive analytics

Background:

• Grey Model (GM(1,1)) prediction accuracy is sensitive to background value estimation.
• Traditional trapezoidal background values have limited applicability to specific data sequences.

Purpose of the Study:

• To propose an improved GM(1,1) background value reconstruction approach.
• To enhance the applicability and prediction accuracy of the GM(1,1) model.

Main Methods:

• Developed Model I: Trapezoidal background value function with adjustable parameters.
• Developed Model II: A novel background value function using parameter sequences.
• Employed genetic algorithms to optimize parameters for both models.

Main Results:

• Model I and Model II demonstrated superior prediction accuracy for exponential growth data compared to traditional methods.
• Model I improved prediction accuracy by 0.3643% and 0.2725% over existing models for road traffic volume data.
• Model II further enhanced prediction accuracy by 0.1075% compared to Model I.

Conclusions:

• The proposed intelligent trapezoidal and variable weight methods significantly improve GM(1,1) model background value estimation.
• The enhanced GM(1,1) models offer greater accuracy and broader applicability for time series forecasting.