关于稀疏最小二乘支持向量回归机的改进剪枝算法

Improved pruning algorithms for sparse least squares support vector regression machine

对于含有奇异点的系统而言,由于一般的剪枝算法不能成功抑制系统中的奇异点,在借鉴支持向量分类机选择支持向量方法的基础上,提出了改进的剪枝算法.改进的剪枝算法能成功抑制系统中的奇异点,减少支持向量的数目,增强最小二乘支持向量回归机的泛化能力.另外,仿真实例也验证了改进剪枝算法的有效性.在不含有奇异点系统中,改进的剪枝算法退化成了一般的剪枝算法,也就是说一般的剪枝算法是改进剪枝算法的一个特例.

For a system with outliers, the common pruning algorithms don＇t oppress outliers. Therefore, the improved pruning algorithms are proposed based on the selecting support vectors method of support vector classifier. The improved pruning algorithms not only oppress outliers but also reduce the number of support vectors and boost the generalization performance of least squares support vector regression machine, and the simulation example corroborates the efficacy of the improved pruning algorithms. Moreover, the improved pruning algorithms can also be applied to a system without outliers, and in this situation, the improved pruning algorithms degenerate into the common pruning algorithms, i. e., the common pruning algorithms are special cases of the improved pruning algorithms.