改进矮猫鼬优化算法的特征选择

Improved Dwarf Mongoose Optimization Algorithm for Feature Selection

将矮猫鼬优化(DMO)算法与学习策略相结合,提出一种新的改进矮猫鼬优化(IDMO)算法。首先,提出随机准反向反射学习(RQORBL)策略并应用于阿尔法组,以提高全局探索能力;其次,引入动态透镜成像反向学习(LOBL)策略,以平衡算法的探索和开发,提升跳出局部最优的能力。为验证新算法性能,将IDMO与几种新近提出的优化算法进行对比,并对UCI存储库中的10个数据集进行特征选择仿真实验。实验结果表明IDMO寻优能力更佳,跳出局部最优能力明显增强,能够有效适用于特征选择问题。

In this paper,a new improved dwarf mongoose optimization(IDMO)algorithm is proposed by combining dwarf mongoose optimization(DMO)with learning strategies.First of all,random quasi opposition reflection based learning(RQORBL)strategy is proposed and applied to alpha group to improve the ability of global exploration.Secondly,a dynamic lens imaging opposition-based learning(LOBL)strategy is introduced to balance the exploration and exploitation of algorithms and improve the ability to jump out of local optimization.To verify the performance of the new algorithm,IDMO is compared with several newly proposed optimization algorithms,and feature selection simulation experiments are conducted on 10 datasets in the UCI(University of California,Irvine)repository.The experimental results show that IDMO has better optimization ability,and its ability to jump out of local optimization is significantly enhanced,which can be effectively applied to feature selection problems.