An Improvement of Bird Swarm Algorithm and Its Applications

There are three behaviors in the two levels original bird swarm algorithm (BSA): forage behavior and alert behavior in first level, migration behavior in the second level. In the improved method, a few executing birds with migration behavior is modified in the first level behaviors. 5 different benchmark functions are optimized to verily the effectiveness of the improved method, the results illustrate that the accuracy and convergence speed are both increased. In addition, a case study of neural network is carried out to optimize the weights and thresholds, and the improved BSA achieves smaller errors than the original BSA.