Chaotic RIME optimization algorithm with adaptive mutualism for feature selection problems.

Area of Science:

• Computational Intelligence
• Optimization Algorithms
• Machine Learning

Background:

• Swarm-based optimization algorithms like RIME face limitations in exploration-exploitation balance, leading to local optima and slow convergence.
• Enhancing the search mechanism is crucial for discovering diverse and optimal solutions in complex problems.

Purpose of the Study:

• To introduce the Adaptive Chaotic RIME (ACRIME) algorithm, designed to overcome the limitations of the original RIME algorithm.
• To improve population diversity, balance exploration and exploitation, and enhance local and global search capabilities.
• To evaluate ACRIME's effectiveness on benchmark functions, real-world feature selection tasks, and COVID-19 classification.

Main Methods:

• Developed ACRIME by incorporating intelligent population initialization using chaotic maps, a modified Symbiotic Organism Search (SOS) mutualism phase, a mixed mutation strategy, and a restart strategy.
• Assessed ACRIME using CEC2005 and CEC2019 benchmark functions.
• Applied ACRIME to fourteen datasets for feature selection and to COVID-19 classification data.
• Compared ACRIME against other metaheuristics using Wilcoxon rank-sum and Friedman rank tests.

Main Results:

• ACRIME demonstrated superior performance and competitiveness against established algorithms.
• The algorithm effectively identified optimal feature subsets, improving classification accuracy while reducing feature count.
• Statistical tests confirmed the significant performance improvements of ACRIME.

Conclusions:

• ACRIME successfully enhances the RIME algorithm's performance by improving the exploration-exploitation balance and search scope.
• The proposed algorithm shows significant potential for real-world applications, particularly in feature selection and classification tasks.
• ACRIME offers a robust and effective approach to complex optimization problems.