磷虾群算法优化支持向量机的威胁估计  

Threat assessment of support vector machine optimized by Krill Herd algorithm

作　　者：黄璇[1,2] 郭立红[1] 李姜[1] 于洋[1] 

机构地区：[1]中国科学院长春光学精密机械与物理研究所,吉林长春130033 [2]中国科学院大学,北京100039

出　　处：《光学精密工程》2016年第6期1448-1455,共8页Optics and Precision Engineering

基　　金：国家自然基金青年基金(No.61205143)

摘　　要：为提高目标威胁估计的预测精度,在传统支持向量机优化方法的基础上,提出了采用磷虾群算法优化支持向量机的威胁估计方法。介绍了磷虾群算法和支持向量机的原理,并基于此采用磷虾群算法对支持向量机中的惩罚参数和核函数参数进行优化,寻找最优的惩罚参数和核函数参数;建立磷虾群优化支持向量机的目标威胁估计模型,并实现基于该模型的目标威胁估计算法。采集90组原始数据组成训练集、30组数据组成测试集,对该目标威胁估计算法进行仿真实验。实验结果显示,磷虾群算法优化支持向量机的预测误差为0.002 91,小于采用粒子群算法或萤火虫算法优化的支持向量机。结果表明,磷虾群优化支持向量机的目标威胁估计方法可以有效地完成目标威胁估计。To put forward the method of threat assessment of support vector machine optimized by Krill Herd algorithm based on the traditional support vector machine optimization method,so as to improve the forecast precision of target threat assessment.The thesis introduces the principles of Krill Herd algorithm and support vector machine and optimize the penalty parameter and kernel function parameter in the support vector machine with Krill Herd algorithm to find the optimal penalty parameter and kernel function parameter;establishes the model of target threat assessment of the support vector machine optimized by Krill Herd algorithm and achieves the target threat assessment algorithm based on this model.Collect 90 sets of original data to form the training set and 30 sets of data to form the test set to carry out simulation experiment on the target threat assessment algorithm.The experimental result shows that the forecast error of the support vector machine optimized by Krill Herd algorithm is 0.002 91 which is less than that of the support vector machine optimized by particle swarm algorithm or firefly algorithm.It can conclusion that the target threat assessment method of the support vector machine optimized by Krill Herd algorithm can effectively complete the target threat assessment.

关 键 词：威胁估计 磷虾群 支持向量机 参数寻优 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]