基于樽海鞘群与粒子群混合优化算法的特征选择

针对现有特征选择方法中存在的收敛速度慢和计算效率低等问题,提出了一种基于樽海鞘群与粒子群优化的混合优化(hybrid optimization of salp swarm algorithm and particle swarm optimization,HOSSPSO)特征选择方法,该方法在樽海鞘群算法(salp swarm algorithm,SSA)的基础上,引入粒子群优化(particle swarm optimization,PSO),提高了SSA的收敛速度,改进了探索和开发步骤的效率,增加了解空间更多的灵活性和多样性,使得方法能够迅速获得全局最优值.为了验证算法的性能,在2个实验序列上进行了测试:第一个实验序列使用基准函数,将HOSSPSO与标准SSA、PSO进行了比较;第二个实验序列采用不同的UCI数据集,通过提出的算法确定最佳特征集.实验结果表明,相比于其他优化算法,HOSSPSO的性能更具优势,在多项评估指标中获得较好的效果,能以极少量的特征获得最大的分类精度.

Feature selection based on hybrid optimization of salp swarm algorithm and particle swarm optimization

Aiming at the problems of slow convergence and low computational efficiency in the existing feature selection methods, we propose a feature selection method based on hybrid optimization of salp swarm algorithm and particle swarm optimization (HOSSPSO). Based on the optimization algorithm of salp swarm algorithm, particle swarm optimization is introduced to improve the convergence speed of SSA algorithm, improve the efficiency of exploration and development steps, and increase the flexibility and diversity of learning space, so that the method can quickly obtain the global optimal value. In order to verify the performance of the algorithm, two experimental sequences are tested. The first experimental sequence is compared with standard SSA and PSO by using benchmark function. At the same time, the second experimental sequence uses different UCI data sets to determine the best feature set through the proposed algorithm. The experimental results show that compared with other optimization algorithms, the proposed algorithm has more advantages in performance, obtains better results in many evaluation indexes, and obtains the maximum classification accuracy with a few features.