Abstract: In the standard Least Squares Support Vector Machine (LS-SVM) for regression, every training sample is equa11y considered, which is unsuitable when there exists significant difference among the training samples. The weighted least squares support vector regression based on AdaBoost algorithm was proposed. Learning by a series of support vector regressions, the proposed approach combined all the results in accordance with some rule. At the same time, adaptive weighted factors in LS-SVM were constructed to control the error function according to the regression error. It emphasized the significant difference among the training samples by adaptive weighted factors and improved the performance of generalization error. The experimental results demonstrate that the proposed approach has a competitive learning ability and acquires better accuracy than LS-SVM.

Key words: AdaBoost algorithm, Least Squares Support Vector Machine (LS-SVM), regression

摘要： 针对二乘向量机(LS-SVM)对所有样本误差惩罚相同、预测精度不高的问题，提出了一种基于AdaBoost模型的二乘向量回归机。该算法使用多个二乘向量机按照某种学习规则协调各二乘向量机的输出，同时根据回归精度，建立各二乘向量机中每一个样本的误差惩罚权重，以突出样本的惩罚差异性，提高算法的泛化性能。实验结果表明，提出的算法提高了二乘向量回归机的预测精度，优化了学习机的性能。

关键词: AdaBoost算法, 二乘向量机, 回归