CAJ | 학술논문

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。
대지지향량궤적핵삼수선취도목전잉몰유형성일투성숙적이론，엄중영향료기엄범적응용。대핵삼수적선취주료일정적탐토。장신경망락여입자군우화산법상결합병용우지지향량궤핵함수적삼수우화。해방법능구동시구유신경망락교강적비선성의합능력화입자군우화산법적심우능력。수치실험결과표명해산법대지지향량궤핵삼수적우화시가행적、유효적，병차구유교고적분류준학솔화교호적추엄성능。
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:com-bining 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.