针对支持向量回归中由于噪声和孤立点带来的过拟合问题,提出了一种基于支持向量数据域描述的加权系数函数模型,根据样本到特征空间最小包含超球球心的距离来确定其加权系数。将提出的加权系数模型用于加权支持向量回归中,一维数据集仿真表明,提出的模型可以有效减小回归误差,提高支持向量回归算法的抗噪声能力。
To overcome the problem of over-fitted caused by noises and outliers in support vector regress(SVR),weighted coefficient model based on support vector data description(SVDD) is used in this paper.The weighted coefficient value to each input sample is confirmed according to its distance to the center of the smallest enclosing hyper-sphere in the feature space.The proposed model is applied to weighted support vector regression(WSVR) for 1-dimensional data set simulation.The results indicate that the proposed method actually reduces the error of regression and yields higher accuracy than support vector regression(SVR) does.