基于加权最小二乘双支持向量机的含噪声分类

摘要

This paper aimed to overcome the influence of the noise on the least squares twin support vector ma-chine. Following the principle that gives the samples containing large noise smaller weights and the samples contai-ning small noise larger weights, the weights reflecting the noise level in samples were first proposed by evaluating the distances from the training samples to the two nonparallel separating hyperplanes. By using the proposed weights, the linear and nonlinear weighted least squares twin support vector machines were presented and the corresponding solving algorithms were developed. The proposed weighted least squares twin support vector machines were applied to the Heart-statlog data and Two-moons data. The simulation results show that the proposed methods can remove the influ-ence of the noise and improve the classification accuracy.%针对最小二乘双支持向量机对噪声样本敏感的问题，依据给含有大噪声的样本赋予较小权重、给较小噪声的样本赋予较大权重的原则，通过评估训练样本点到两个非平行分类超平面的距离，构造了能反映样本噪声程度的权重，提出了线性和非线性加权最小二乘双支持向量机，并发展了两种加权支持向量机的求解算法，解决了对含噪声样本的高精度分类问题。将所提两种加权最小二乘双支持向量机分别应用到Heart-statlog和Two-moons数据集上进行仿真，结果表明所提方法有效消除了噪声的影响，提高了分类精度。