An Improved Artificial Fish-Swarm Algorithm and Its Applications

The convergence and accuracy of the artificial fish swarm algorithm (AFSA) depend not only on the algorithm parameters, but also on the four different behaviors of the fish swarm algorithm. This paper analyzes the influence mechanism of fish swarm algorithm's four behaviors: preying, swarming, following, and random moving behavior on the convergence speed and performance of the algorithm. Combined with the interval division of the bird swarm algorithm, the number of fish swarm of the four behaviors in the AFSA is modified, and a novel algorithm based on interval division (DAFSA) is proposed. Simulations in 5 test functions show that the DAFSA algorithm can quickly converge to a better global solution, has good convergence accuracy. Meanwhile, DAFSA has good practical value in solving the parameters of neural networks.