An ensemble algorithm based on adaptive chaotic quantum-behaved particle swarm optimization with weibull distribution and hunger games search and its financial application in parameter identification
Hanqiu Ye,Jianping Dong
DOI: https://doi.org/10.1007/s10489-024-05537-4
IF: 5.3
2024-05-30
Applied Intelligence
Abstract:Quantum-behaved Particle Swarm Optimization (QPSO) is a meta-heuristic optimization algorithm, which is widely used in many research fields and practical problems due to its flexibility and low computational cost. However, the existing QPSO algorithms and their variants still have problems such as insufficient search capabilities, lack of adaptivity, and prone to stagnation. This paper proposes a novel ensemble algorithm, ACQPSOW-HGS, based on Quantum-behaved Particle Swarm Optimization (QPSO) and Hunger Games Search (HGS). By combining three improvements and introducing three hybrid strategies, our algorithm has made a comprehensive development, effectively improving the stability and solution accuracy in a large number of test functions and the parameter identification application, which is superior compared with many existing algorithms. First, we design the Weibull distribution random number generation operator, the distance-guided adaptive control technique, and the chaotic update mechanism to deal with the weak randomness, insufficient adaptability, and susceptibility to stagnation of the original QPSO algorithm, respectively. Integrating the above improvements, ACQPSOW is proposed as an improved variant of QPSO. Second, the proposed ensemble algorithm ACQPSOW-HGS is built on ACQPSOW and HGS and combined with specific hybrid strategies to add population diversity and improve search efficiency, including the Selection-Crossover-Mutation mechanism, the elite local search mechanism, and the information exchange mechanism. Finally, the experiments, on 23 benchmark functions and the IEEE CEC 2017 test suite, demonstrate that ACQPSOW-HGS outperforms comparison algorithms in terms of convergence speed and solution accuracy through non-parametric statistical tests. Moreover, ACQPSOW-HGS was applied to the fractional-order hyper-chaotic financial system for parameter identification to illustrate the applicability and robustness in solving real-world problems.
computer science, artificial intelligence