摘要
一类支持向量机是只有正类样本的一类分类算法,该算法已经在孤立点检测、经济预警中有了广泛的应用。根据一类分类方法,本文提出一种非线性回归算法,该算法揭示了一类分类、二类分类以及回归之间的关系。该方法首先对训练数据的响应变量向上和向下移动ε,进而获得两个样本集合;然后应用核映射方法在高维特征空间中分别求包含两个集合的最小超球体中心;最后,通过求平分两个中心的间隔最大超平面获得回归函数。两个仿真实验结果验证了所给算法的有效性和可行性。
A new method for nonlinear regression is proposed based on one-classification in this paper,which can uncover the relation among one-class classification,binary classification and regression.The method first obtains two sets based on the training data with the response variable shifted up and down by ε,and then,computes respectively a sphere shaped decision boundary with minimal volume around every set of objects,finally gets the regression function by computing the max-margin which separates the plane and bisects the two centers.The results of two simulation experiments show that the proposed method is feasible and valid.
出处
《计算机工程与科学》
CSCD
北大核心
2012年第7期150-153,共4页
Computer Engineering & Science
基金
辽宁省高等学校科研项目(2008343)
关键词
非线性回归
核映射
一类分类
超平面
nonlinear regression
kernel mapping
one-class classification
hyper plane
