基于神经网络的粒子群算法优化SVM参数问题  

Question of SVM kernel parameter optimization with particle swarm algorithm based on neural network

作　　者：李磊[1] 高雷阜[1] 赵世杰[1] 

机构地区：[1]辽宁工程技术大学理学院,辽宁阜新123000

出　　处：《计算机工程与应用》2015年第4期162-164,259,共4页Computer Engineering and Applications

摘　　要：对支持向量机的核参数选取到目前仍没有形成一套成熟的理论,严重影响了其广泛的应用。对核参数的选取做了一定的探讨。将神经网络与粒子群优化算法相结合并用于支持向量机核函数的参数优化。该方法能够同时具有神经网络较强的非线性拟合能力和粒子群优化算法的寻优能力。数值实验结果表明该算法对支持向量机核参数的优化是可行的、有效的,并且具有较高的分类准确率和较好的推广性能。The selection of Support Vector Machine（SVM）of the kernel parameter is still not formed a set of mature theory,thus affecting its widespread application seriously. This paper makes a discussion on the kernel parameter selection：combining neural network with particle swarm optimization algorithm for application in the parameters of the kernel function of Support Vector Machine. This method possesses strong fitting ability of nonlinear neural network and searching capability of the particle swarm optimization algorithm. Numerical experiment results show that the algorithm of Support Vector Machine nuclear parameter optimization is feasible, effective, and has higher classification accuracy and good generalization performance.

关 键 词：支持向量机 粒子群优化算法 参数优化 神经网络 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]