基于随机邻域变异和趋优反向学习的差分进化算法.

The traditional differential evolution（DE） algorithm balanced global exploration and local exploitation inadequately, and had problems with easily falling into local optimal solutions, low solution accuracy and slow convergence speed. Therefore, this paper proposed a differential evolution algorithm based on random neighborhood mutation and optimal opposition-based learning（RNODE） and analyzed for its complexity. Firstly, the algorithm generated a random neighborhood for each individual in the current population, and used the global best individual to guide the neighborhood best individual to generate a composite basis vector, combined with an adaptive update mechanism of the control parameters to constitute a random neighborhood mutation strategy, which enabled the algorithm maintained its exploration ability and guided the population towards the optimal direction. Secondly, to further help the algorithm jump out of the local optimum, the algorithm performed the optimal opposition-based learning strategy on the poorer individuals to expand the search area. Finally, this paper compared RNODE with 9 algorithms to verify the effectiveness and advancement of RNODE. The experimental results on 23 benchmark functions and 2 real-world engineering optimization problems show that the RNODE algorithm has a higher convergence accuracy, faster speed and a greater stability. [ABSTRACT FROM AUTHOR]