Improved Artificial Ecosystem Optimization For Numerical Optimization Problems

Artificial ecosystem based optimization(AEO),as a popular meta-heuristic algorithm,has been widely used in various practical optimization problems.However,it still has many problems,such as: unable to balance exploration and exploitation well,easy to fall into local optimum,and loss of population diversity in the late iteration.In order to improve these problems,further improve the performance of the original AEO,the following work is done in this thesis:(1)In order to better balance AEO exploration and exploitation and improve population diversity,Muti-subgroup artificial ecosystem optimization with comprehensive learning(MCLAEO)is proposed.In MCLAEO,firstly,the improved comprehensive learning is applied to AEO so that the individuals do not have to learn all the dimensions of a particle,thus improving the diversity of the population.Secondly,using an improved Lévy flight to generate to reduce step size and improve exploitation.At the same time,the consumer part is divided into three dynamic subgroups,each of which represents a different level of individuals,and all of which are focused on exploration or exploitation,so as to better balance global exploration and local exploitation.Finally,MCLAEO is compared with eight other well-known evolutionary algorithms on 29 benchmark functions of CEC2017.The experimental results and statistical analysis show that the proposed MCLAEO has good performance.(2)In order to further improve the performance of the algorithm,An improved artificial ecosystem optimization(AIAEO)is proposed.In AIAEO,firstly,a new initialization method is proposed by simulating the refraction law of light to improve the quality of the initial solution.Compare two random numbers and decide whether to do incident initialization or refraction initialization.Second,Lévy flight and Brownian motion are combined to better balance global exploration capability and local exploitation capability.Third,add Cauchy operator and cosine operator to increase disturbance and avoid falling into local optimum.Finally,the proposed AIAEO is compared with the original AEO and seven other popular metaheuristic algorithms on 29 benchmark functions of CEC2017,and the effectiveness of the proposed AIAEO is verified.