关键词: 群体智能, 人工蜂群算法, 动态评价选择策略, 收敛精度, 函数优化

Abstract:

To overcome the problem of easily trapping into local optima of standard Artificial Bee Colony (ABC) algorithm, the roulette selection strategy of ABC was modified and an improved ABC based on dynamic evaluation selection strategy (DSABC) algorithm was proposed. Firstly, the quality of each food source position was evaluated dynamically according to the times that the food source position had been continuously updated or stagnated within a certain number of iterations so far. Then, onlooker bees were recruited for the food source according to the obtained value of the evaluation function. The experimental results on six benchmark functions show that, compared with standard ABC algorithm, the proposed dynamic evaluation selection strategy modifies the selection strategy of ABC algorithm, and greatly improves the quality of solution of DSABC algorithm, especially for function Rosenbrock with different dimensions, the absolute error of the best solution reduces from 0.0017 and 0.0013 to 0.000049 and 0.000057, respectively; Moreover, DSABC algorithm can avoid the premature convergence caused by the decrease of population diversity at later stage and improve the accuracy and stability of solutions, thus provides an efficient and reliable solution method for function optimization.

Key words: swarm intelligence, Artificial Bee Colony (ABC) algorithm, dynamic evaluation selection strategy, convergence accuracy, function optimization