| 基于非线性空间的小样本回归方法 |
| |
| 引用本文： | 于哲夫,路慧彪,贾传荧. 基于非线性空间的小样本回归方法[J]. 大连海事大学学报(自然科学版), 2010, 0(4) |
| |
| 作者姓名： | 于哲夫 路慧彪 贾传荧 |
| |
| 作者单位： | 大连海事大学航海学院,辽宁,大连,116026;大连海事大学交通与物流工程学院,辽宁,大连,116026 |
| |
| 摘 要： | 为解决小样本回归时引起的过学习问题并提高回归精度,提出一种基于非线性空间特征选择的支持向量机.该方法依据矩阵相似度量或从研究的实际问题出发,绕过核技巧,直接将原始输入空间映射为适宜的非线性空间.该方法运用遗传算法在维数较多的非线性空间中搜索对输出影响最大的一些特征,达到降低输入空间维数的目的,从而避免过学习问题,并可获得简明的非线性回归函数.
|
| 关 键 词： | 非线性空间 特征选择 支持向量回归机 遗传算法(GA) |
Nonlinear space-based regression suitable for small samples |
| |
| YU Zhefua,LU Huibiaob,JIA Chuanyinga. Nonlinear space-based regression suitable for small samples[J]. Journal of Dalian Maritime University, 2010, 0(4) |
| |
| Authors: | YU Zhefua LU Huibiaob JIA Chuanyinga |
| |
| Affiliation: | YU Zhefua,LU Huibiaob,JIA Chuanyinga(a.Navigation College,b.Transportation , Logistics Engineering College,Dalian Maritime University,Dalian 116026,China) |
| |
| Abstract: | For solving the problem of over-fitting caused by small samples and obtaining better regression accuracy,a support victor machine based on feature extraction from nonlinear space was proposed.The original input features space was mapped to nonlinear space based on matrix similarity,and genetic algorithm was used to extract features from the nonlinear space to realize the purpose of reducing dimensions of input space,which can avoid over-fitting,and get a concise nonlinear regression function. |
| |
| Keywords: | nonlinear space feature extraction support vector regression genetic algorithm(GA) |
| 本文献已被 CNKI 万方数据 等数据库收录！ |
|
