摘要
针对基本灰狼优化算法(grey wolf optimizer,GWO)在求解复杂优化问题时存在解精度低、探索与开发能力不平衡、收敛速度慢和易陷入局部最优的缺点,提出一种基于多策略融合的改进灰狼优化算法。首先,设计一种基于正弦函数的非线性过渡参数策略替代原灰狼优化算法中的线性递减策略,以实现算法从勘探到开发的良好过渡;其次,利用个体自身历史最佳位置和决策层个体共同引导群体进行搜索,以加速算法收敛速度和提高寻优精度;然后,在当前最优灰狼个体上引入小孔成像学习策略产生新的候选个体,以降低算法陷入局部最优的概率。选取6个基准测试函数进行数值实验。结果表明:改进算法在求解精度和收敛速度指标上均优于其他比较算法。最后,将改进算法用于求解特征选择问题,对10个基准数据集的仿真结果表明,改进算法能有效地提高分类精度和选择最优特征。
Aiming at the shortcomings of the basic grey wolf optimizer(GWO)algorithm such as low solution,unbalanced between exploration and exploitation,slow convergence speed and easy to fall into local optima,an improved GWO algorithm integrated with multiple-strategies was proposed.Firstly,a nonlinear transition parameter strategy based on sine function was designed to replace the original linear decrease strategy of the basic GWO algorithm,which achieved a good transition from the exploration to exploitation.Secondly,the grey wolf personal historical best position and the decision-making level individuals were simultaneously used to guide the search of the other grey wolves,which accelerated the convergence speed and improved the solution precision.Then,the pinhole-imaging learning strategy was introduced on the current optimal grey wolf individual to generate the new candidate individual,which reduced the probability of falling into local optima.To verify the effectiveness of the proposed algorithm,six benchmark test functions from literature were used.The experimental results show that the proposed algorithm is better than the other compared methods in terms of solution precision and convergence speed.Finally,the proposed algorithm was applied to solve the feature selection problems.The simulation results on 10 benchmark data sets demonstrate that the proposed algorithm can effectively enhance the classification accuracy and select the optimal features.
作者
徐明
龙文
XU Ming;LONG Wen(Guizhou Key Laboratory of Big Data Statistics Analysis,Guizhou University of Finance&Economics,Guiyang 550025,China;School of Mathematics and Statistics,Guizhou University of Finance&Economics,Guiyang 550025,China;Guizhou Key Laboratory of Economics System Simulation,Guizhou University of Finance&Economics,Guiyang 550025,China)
出处
《科学技术与工程》
北大核心
2021年第20期8544-8551,共8页
Science Technology and Engineering
基金
国家自然科学基金(61463009)
贵州省科学技术基金(黔科合基础[2020]1Y012)
贵州省大数据统计分析重点实验室开放课题(BDSA20190106,BDSA20200101)
贵州省教育厅创新群体项目(黔教合KY字[2021]015)。
关键词
灰狼优化算法(GWO)
特征选择
函数优化
多策略
反向学习
grey wolf optimizer algorithm(GWO)
feature selection
function optimization
multiple-strategies
opposition learning
作者简介
第一作者:徐明(1976-),男,汉族,湖北荆州人,博士,教授。研究方向:数据挖掘及智能优化算法。E-mail:xuming@mail.gufe.edu.cn;通信作者:龙文(1977-),男,汉族,湖南隆回人,博士,教授。研究方向:智能优化算法、数据挖掘及工程优化。E-mail:lw227@mail.gufe.edu.cn。
