Multi-Strategy Adaptive Cuckoo Search Algorithm

Cuckoo search algorithm (CS) is a powerful biological-inspired search algorithm, which is widely used in continuous space optimization problems. However, a single search strategy in CS makes all cuckoos have similar search behavior, and it is liable to plunges into local optimal. In addition, whether CS can successfully solve a problem largely depends on the value of control parameters. Using the trial and error method to determine the value of parameters will cost a lot of computational expense frequently. In order to solve these problems, a multi-strategy adaptive cuckoo algorithm (MSACS) is proposed in this paper. Firstly, five search strategies are adapted to cooperate with each other, and the use of various previous strategies and control parameters are studied. The probability of each strategy being used and the value of control parameters are changed adaptively. Then, the performance of MSACS is tested and evaluated on 24 common benchmark functions. Finally, several advanced CS algorithms, particle swarm algorithm (PSO) and differential evolution algorithm (DE) variants will be compared with MSACS. The results show that the MSACS is better than the algorithms above.