摘要
在实际应用中支持向量机的参数选取问题一直没有得到很好地解决,限制了其应用。为了能够自动获取最优的支持向量机参数,提出了基于模糊遗传算法的SVM参数选择方法,用模糊逻辑在线调整遗传算法的交叉概率pc和变异概率pm,并采用基于模糊遗传优化的支持向量机回归和BP神经网络对非线性系统辨识问题进行了研究。仿真结果表明,在小样本情况下,支持向量机比神经网络具有更高的系统辨识精度和更好的泛化能力。
In practice, the problem on how to select parameters of SVM is not solved properly,so that its application is restricted. In order to get the optimal parameters of SVM automatically, a parameter selection approach based on fuzzy genetic algorithms （FGAs） is proposed in this paper. The nonlinear system identification is studied using the crossover probability Pc and mutation probability Pm of the on-line adjustment genetic algorithm based on fuzzy logic, the support vector regression based on FGAs and BP neural network. The simulation results show that SVM has higher system identification precision and better generalization performance than neural networks in the case of small samples.
出处
《西安理工大学学报》
CAS
2006年第1期49-53,共5页
Journal of Xi'an University of Technology
基金
陕西省自然科学基金资助项目(2003F33)
陕西省教育厅专项科研项目(04JK249)
关键词
支持向量机
模糊遗传算法
结构风险最小化
神经网络
系统辨识
support vector machine
fuzzy genetic algorithms （FGAs）
structural risk minimization
neural network
system identification
