Boosting Slime Mould Algorithm for High-Dimensional Gene Data Mining: Diversity Analysis and Feature Selection.

Area of Science:

• Computational Intelligence
• Optimization Algorithms
• Bio-inspired Computing

Background:

• The Slime Mould Algorithm (SMA) is a metaheuristic inspired by slime mould foraging behavior.
• The basic SMA can struggle with complex problems, potentially leading to local optima.
• There is a need for improved algorithms to enhance global search capabilities and avoid premature convergence.

Purpose of the Study:

• To propose an improved Slime Mould Algorithm (SMA) that addresses the limitations of the basic version.
• To enhance the global search performance and population diversity of the SMA.
• To validate the effectiveness of the improved algorithm in continuous optimization and discrete feature selection tasks.

Main Methods:

• Introduced Gaussian mutation to increase population diversity.
• Incorporated Levy flight to mitigate local optima and improve global search.
• Developed continuous (GLSMA) and discrete (BGLSMA) versions of the enhanced algorithm.
• Tested GLSMA on 33 classical continuous optimization problems.
• Validated BGLSMA on 14 high-dimensional gene datasets for feature selection.

Main Results:

• The continuous version (GLSMA) demonstrated improved performance over the original SMA, alleviating its defects.
• The discrete version (BGLSMA) achieved higher classification accuracy with fewer selected features in high-dimensional gene datasets.
• Experimental results confirmed the enhanced global search capability and reduced susceptibility to local optima.

Conclusions:

• The proposed enhancements effectively improve the Slime Mould Algorithm's performance.
• The improved algorithm, GLSMA/BGLSMA, shows significant practical value, particularly in high-dimensional gene feature selection.
• Gaussian mutation and Levy flight are effective strategies for optimizing metaheuristic algorithms like SMA.