摘要
针对鲸鱼优化算法容易陷入局部最优,求解精度低,收敛速度慢,提出了一种多种群进化和差分变异的鲸鱼优化算法(MDWOA).首先,根据适应度值将种群划分为两个大小相等的子种群,并为每个子种群分配不同的移动策略,以平衡全局和局部搜索能力.其次,设计了一种种群进化和差分变异的策略来帮助MDWOA提高收敛速度,避免其陷入局部最优.最后,引入反向学习策略,增加种群多样性.将MDWOA与多种优化算法在13个基准函数上进行仿真测试,非参数检验的结果表明相较于其他优化算法来说改进的算法具有更高的精度和稳定性.在此基础上,建立了基于MDWOA优化BP神经网络模型,预测波士顿房价的实验结果表明所提出的预测模型具有更好的预测性能和有效性.
Aiming at the disadvantages of whale optimization algorithm,such as easily falling into the local optimum,low accuracy and slow convergence speed,a whale optimization algorithm with multi-population evolution and differential mutation is proposed.Firstly,individual whales are classified into two equally sized subpopulations according to their fitness values,and different movement strategies are assigned to each subpopulation in order to balance global and local searching capabilities.Secondly,a strategy of population evolution and differential mutation is designed to assist MDWOA in improving its convergence speed and avoiding getting trapped in local optima.Finally,opposition-based learning strategy is introduced to increase the diversity of the population.Simulation tests of MDWOA and multiple optimization algorithms are conducted on 13 benchmark functions,where the non-parametric test results indicate that MDWOA performs better in accuracy and stability compared to other optimization algorithms.Based on this,a BP neural network prediction model based on MDWOA is established.And the experimental results of predicting Boston house prices show that the proposed prediction model exhibits better predictive performance and effectiveness.
作者
朱杰
付伟
马宁
季伟东
苏婷
陈珊
ZHU Jie;FU Wei;MA Ning;JI Weidong;SU Ting;CHEN Shan(Harbin Normal University School of Computer Science&Information Engineering,Harbin 150025,China)
出处
《小型微型计算机系统》
CSCD
北大核心
2024年第11期2618-2627,共10页
Journal of Chinese Computer Systems
基金
黑龙江省自然科学基金项目(LH2021F037)资助
黑龙江省高等教育教学改革项目(SJGY20210455)资助
哈尔滨市科技局科技创新人才研究专项项目(2017RAQXJ050)资助
哈尔滨师范大学博士科研启动基金项目(XKB201901)资助
哈尔滨师范大学计算机学院科研项目(JKYKYY202006)资助
哈尔滨师范大学研究生培养质量提升工程项目(HSDYJSJG2019006)资助
哈尔滨师范大学计算机科学与信息工程学院自然科学基金项目(JKYKYY202102)资助.
关键词
多种群进化
差分变异
鲸鱼优化算法
反向学习
MDWOA-BP神经网络
multi-population evolution
differential mutation
whale optimization algorithm
opposition-based learning
MDWOA-BP neural network
