Research on residual GM optimization based on PEMEA-BP correction.

Area of Science:

• Computational Intelligence
• Data Modeling
• Machine Learning

Background:

• Grey Model (GM) offers advantages for small sample sizes and high accuracy.
• GM faces limitations including high input data demands and significant error margins.
• Existing algorithms often suffer from slow convergence and local minima issues.

Purpose of the Study:

• To propose an advanced algorithm for modifying the Grey Model (GM) residual tail.
• To enhance the applicability of GM to non-linear and multidimensional datasets.
• To overcome the limitations of traditional GM and other machine learning algorithms.

Main Methods:

• Development of a novel algorithm integrating Populational Entropy Based Mind Evolutionary Algorithm with Error Back Propagation Training Artificial Neural Algorithm.
• Application of the algorithm to modify the residual tail of the Grey Model (GM).
• Comparative analysis using statistical indicators such as SSE, MAE, MSE, MAPE, and MRE.

Main Results:

• The proposed algorithm significantly improves simulation accuracy over standard GM, BP, and combined genetic algorithms.
• Demonstrated faster convergence speeds in small sample data experiments.
• Successfully expanded the scope of GM for non-linear and multidimensional objects.

Conclusions:

• The new algorithm retains GM's strengths while addressing its core weaknesses.
• It offers a robust solution for complex modeling tasks with limited data.
• This approach provides a more accurate and efficient alternative to existing methods.