A hybrid binary dwarf mongoose optimization algorithm with simulated annealing for feature selection on high...

Area of Science:

• Computational Intelligence
• Machine Learning
• Optimization Algorithms

Background:

• The Dwarf Mongoose Optimization (DMO) algorithm, while effective for exploration, has limitations in its exploitation phase for engineering design problems.
• Existing metaheuristic approaches require enhancement to balance exploration and exploitation for optimal performance.

Purpose of the Study:

• To introduce a novel hybrid algorithm, BDMSAO, by integrating the binary variant of DMO (BDMO) with Simulated Annealing (SA).
• To improve the exploitative capabilities of the BDMO algorithm for enhanced performance in optimization tasks.

Main Methods:

• Developed the Binary Dwarf Mongoose Optimization-Simulated Annealing (BDMSAO) hybrid algorithm.
• Utilized BDMO as the global search mechanism and SA as the local search component.
• Evaluated the algorithm on 18 UCI machine learning datasets and 3 high-dimensional medical datasets for feature selection.

Main Results:

• The BDMSAO algorithm demonstrated superior performance in feature selection tasks compared to ten other methods.
• Achieved an overall classification accuracy of 61.11%, with 100% accuracy on 9 out of 18 datasets.
• Obtained maximum accuracy on high-dimensional medical datasets and showed competitive feature selection efficiency.

Conclusions:

• The BDMSAO hybrid algorithm effectively addresses the limitations of the original DMO, particularly in the exploitation phase.
• BDMSAO proves robust and efficient for challenging feature selection problems across datasets of varying dimensions.
• The proposed method offers a promising advancement in metaheuristic optimization for machine learning applications.