| 基于遗传算法的SVM参数组合优化 |
| |
| 引用本文： | 刘鲭洁,陈桂明,刘小方,杨庆.基于遗传算法的SVM参数组合优化[J].计算机应用与软件,2012,29(4):94-96,100. |
| |
| 作者姓名： | 刘鲭洁 陈桂明 刘小方 杨庆 |
| |
| 作者单位： | 第二炮兵工程学院504教研室,陕西西安,710025 |
| |
| 基金项目： | 国防预研项目(403030101) |
| |
| 摘 要： | 核函数类型、核函数参数及错误惩罚因子是影响SVM学习能力和泛化能力的关键因素.实际应用中选择上述SVM参数组合多依赖经验或人工尝试,通常很难选择到最优参数组合.提出一种基于遗传算法的SVM优化技术,针对优化对象设计二进制编码基因串和相应遗传算子,能够实现同时对上述三个参数组合的优化.在UCI标准数据库上的实验结果说明了提出方法的有效性.
|
| 关 键 词： | 支持向量机 核函数 参数选择 编码 遗传算法 |
GENETIC ALGORITHM BASED SVM PARAMETER COMPOSITION OPTIMIZATION |
| |
| Liu Qingjie
,
Chen Guiming
,
Liu Xiaofang
,
Yang Qing.GENETIC ALGORITHM BASED SVM PARAMETER COMPOSITION OPTIMIZATION[J].Computer Applications and Software,2012,29(4):94-96,100. |
| |
| Authors: | Liu Qingjie Chen Guiming Liu Xiaofang Yang Qing |
| |
| Affiliation: | Liu Qingjie Chen Guiming Liu Xiaofang Yang Qing(Staff Room 504,The Second Artillery Engineering College,Xi’an 710025,Shaanxi,China) |
| |
| Abstract: | Kernel function type,kernel function parameter and error penalty factor are essential factors that influence SVM learning capability and generalization capability.In real applications,since it usually relies on experience or manual attempts to choose the above mentioned SVM parameter composition,it is usually hard to find an optimum parameter composition.A GA-based SVM optimization technology is proposed.It concentrates on the optimized object to design a binary chromosome string and its corresponding genetic operators,so that it becomes possible to simultaneously realize the optimization of the composition of the above three parameters.Experiment results on UCI standard databases illustrate the validity of the proposed approach. |
| |
| Keywords: | Support vector machine(SVM) Kernel function Parameter selection Coding Genetic algorithm(GA) |
| 本文献已被 CNKI 万方数据 等数据库收录！ |
|
