摘要
介绍了最小二乘支持向量机(LS-SVM)的基本原理,引入了一种混合核函数,讨论了各种参数对LS-SVM的影响。在此基础上,采用改进粒子群算法(IPSO)对惩罚参数、核参数和调节参数进行了优化选取,提高了LS-SVM的学习能力和推广能力。通过实例计算验证了IPSO优化LS-SVM参数的有效性,优化LS-SVM具有较高的GPS高程拟合精度。
The basic theory of least squares support vector machine （LS-SVM） is presented and a mixtures of kernels is introduced. Then, the influence of various parameters included penalty parameter, kernel parameter and adjustable parameter on LS-SVM is discussed. On this basis, a method is put forward by using improved particle swarm optimization （IPSO） to optimize the parameters of LS-SVM which has better global convergence and robustness. The testing example validates the parameters optimized IPSO are effective and LS-SVM base on IPSO has better accuracy in GPS elevation fitting.
出处
《工程勘察》
2013年第7期76-78,共3页
Geotechnical Investigation & Surveying
关键词
粒子群算法
最小二乘支持向量机
高程拟合
混合核函数
particle swarm optimization
least squares support vector machine
elevation fitting
mixtures of kernels
