Research on Adaptive Particle Swarm Optimization Algorithm Based on Diversity-Driven Optimization

• Published in: 2024 5th International Conference on Artificial Intelligence and Electromechanical Automation (AIEA) pp. 1039 - 1042
• Main Authors: Wang, Peishuo, Zuo, Jialiang, Zhang, Zhihao, Lei, Pengfei
• Format: Conference Proceeding
• Language: English
• Published: IEEE 14.06.2024
• Subjects: Boosting; Classification algorithms; Convergence; DAAPSO algorithm; Diversity driven; Diversity reception; Feature extraction; linearly decreasing inertia weight; Optimization; Particle swarm optimization; Smoothing methods; Splines (mathematics)
• DOI: 10.1109/AIEA62095.2024.10692728

A novel Diversity Actuate Adaptive Particle Swarm Optimization (DAAPSO) algorithm is introduced to enhance convergence speed and address premature convergence challenges in particle swarm optimization. Integrating an inertia weight linear decline mechanism with a diversity-driven speed strategy, DAAPSO effectively balances exploration and exploitation. Empirical studies demonstrate the superior performance of DAAPSO across various dimensions compared to existing algorithms. Specifically, experiments on curve energy smoothing and feature selection indicate that DAAPSO significantly improves the smoothness of B-spline curves, reduces curvature peaks post-smoothing, and enhances the accuracy of 5nearest neighbor classifiers when compared to alternative feature selection methods. These results highlight the robust optimization capabilities of DAAPSO, showcasing its effectiveness in both benchmark functions and real-world applications. The findings underscore the efficiency, practicality, and broad applicability of the DAAPSO algorithm.