Stability analysis of the human behavior-based particle swarm optimization without stagnation assumption
Hao Liu,XuWei Zhang,Hong Liang,LiangPing Tu
DOI: https://doi.org/10.1016/j.eswa.2020.113638
IF: 8.5
2020-11-01
Expert Systems with Applications
Abstract:<p>Human behaviour-based particle swarm optimization (HPSO) is an improved particle swarm optimization (PSO) algorithm, which can effectively increase the diversity of population, convergence speed and precision by changing particles' flight schema. In this paper, the first- and second-order stabilities of HPSO on the assumption that the personal optimum, the global optimum and the global worst particle are mutual independence during the whole evolutionary process, while the expectations and variances of particles' positions vary with time so that our calculations do not require the stagnation assumption. The first- and second-order stabilities of HPSO are expressed by stochastic recurrence relation. The convergence boundaries of HPSO are obtained, and which show that the stabilities of HPSO are independent of the expectations and variances of the personal best and the global best. In addition, the reason for choosing the coefficient of the impelled/penalized term in HPSO is analysed, and the results of simulation experiments further confirm the theoretical analysis. Finally, HPSO is used to solve four real-world engineering design problems. Experimental results further confirm the theoretic results and illustrate that the performance of HPSO is promising and competitive.</p>
computer science, artificial intelligence,engineering, electrical & electronic,operations research & management science