融合Levy飞行和精英反向学习的WOA-SVM多分类算法

元启发算法-SVM是多分类评价模型的典型架构,在多分类综合决策判定中具有重要的理论与实践意义,为此提出了一种融合Lévy飞行和精英反向学习的鲸鱼优化算法(Lévy flight and elite opposition-based whale optimization algorithm,LFEO-BWOA)-SVM多分类评价算法.利用Lévy飞行策略替代螺旋轨迹策略更新位置信息,有效克服了鲸鱼优化算法易陷入局部寻优的不足;引入精英反向学习机制增加种群多样性,提高了鲸鱼优化算法全局寻优的能力.实验仿真结果表明,LFEO-BWOA-SVM算法在分类准确率上比传统SVM、BP神经网络分别提高17.84％和4.51％,准确率为98.73％,在训练时间上比标准WOA-SVM和PSO-SVM分别缩短了9.34％和84.94％.实验结果证明,LFEO-BWOA-SVM算法的寻优能力和收敛速度均有明显提升,准确率和快速性良好.

A multi-class algorithm of WOA-SVM using Levy flight and elite opposition-based learning

Meta-heuristic algorithm-SVM is a typical framework of multi-classification evaluation model, which has important theoretical and practical significance in multi-classification comprehensive decision-making. Therefore, this paper proposed a multi-classification evaluation algorithm of an improved whale optimization algorithm using Levy flight and elite opposition-based learning SVM(LFEO-BWOA-SVM). It used Levy flight strategy instead of spiral trajectory strategy to update position information effectively, and could overcome the deficiency of whale optimization algorithm which was easy to fall into local optimization. It introduced the elite opposition-based learning mechanism to increase population diversity and improved the global optimization ability of whale optimization algorithm. The simulation results show that the classification accuracy of LFEO-BWOA-SVM algorithm is 17.84% and 4.51% higher than that of traditional SVM and BP neural network, the accuracy is 98.73%, and the training time is 9.34% and 84.94% shorter than that of standard WOA-SVM and PSO-SVM, respectively. The experimental results show that the optimization ability and convergence speed of LFEO-BWOA-SVM algorithm are obviously improved, and the accuracy and rapidity are good.