A Novel Social Opinion Dynamics Guided Particle Swarm Optimization

In society, the mutual influence and interaction between individuals constitute a social network, and opinion dynamics studies the generation, diffusion and aggregation of thoughts or behaviors in social networks. This paper introduces the idea of evolution in opinion dynamics models into particle swarm optimization algorithm, and proposes a social opinion dynamics-guided particle swarm optimization algorithm (SODPSO). Firstly, in the process of population evolution, the idea of dynamic bounded confidence is used to select the learning object (the best individual in the confidence bound) for each particle to update, and for the individual whose learning object is itself, a difference operator is introduced to update it. Secondly, when the population stagnation reaches a certain threshold, the concepts of individual differences and acceptance are introduced. The particles are sorted and classified according to the fitness value, and different evolution strategies are used to update them in order to jump out of the current optimal solution. Finally, this paper compares SODPSO with the other five PSO variants on part of cec'17 benchmark functions. The experimental results demonstrate that the SODPSO proposed in this paper has greater advantages in functions with certain specific characteristics.