基于支持向量回归的单步多分类算法

分析了利用支持向量回归求解多分类问题的思想,提出了一种基于局部密度比权重设置模型的加权最小二乘支持向量回归模型来单步求解多分类问题：该方法先分别对类样本中每类样本利用局部密度比权重设置模型求出每个样本的权重隶属因子,然后运用加权最小二乘支持向量回归算法对所有样本进行训练,获得回归分类器。为验证算法的有效性,对UCI三个标准数据集以及一个随机生成的数据集进行实验,对比了多种单步求解多分类问题的算法,结果表明,提出的模型分类精度高,具有良好的鲁棒性和泛化性能。

One-step multi-classification algorithm based on support vector regression

A one-step multi-classification algorithm is proposed based on weighted least squares support vector Machine(WLS-SVM) in this paper.A Local-Density-Ratio(LDR) model is applied to weight-setting strategy in the WLS-SVM.In the proposed algorithm,we firstly assign weight membership to each sample in every category by LDR model.Then,we train the whole training set to obtain the regression classifier by WLS-SVM.Numerical experiments are carried out on three benchmarking datasets and one randomly generated dataset.Compared with other methods,the performance of the proposed algorithm is excellent for improving the predicting accuracy of the multi-classification problems.