一种改进的最小二乘孪生支持向量机分类算法

Improvement on Least Squares Twin Support Vector Machine for Pattern Classification

提出了一种新的模式分类器,即广泛权重的最小二乘孪生支持向量机.该支持向量机在正、负两类样本上广泛地增加权重,很好地抑制了交叉噪声样本对数据分类的影响.其次,根据间隔最大化原理,该支持向量机在目标函数上增加了一个正规化项,实现结构风险最小化和避免在求解该目标函数时可能对病态矩阵求逆的处理.同时,提出了利用一种指数函数计算训练样本的密度来获得样本权重值的算法.该算法能够有效缩减计算权重的时间,且具有较强的鲁棒性.实验证明本文提出的广泛权重的最小二乘孪生支持向量机能够实现高精度和高效率的分类效果,而且特别适合于含有交叉噪声样本的数据集分类.

Widely weighted least squares twin support vector machine （WWLSTSVM） is proposed for pattem classification. In WWLSTSVM, weights are widely added on error variables of data samples both in one class and the other. This widely weighted method is especially effective on eliminating the interference of intererossing noise samples. Moreover, a regularizafion term is added with the theory of maximizing margin, in which the structural risk is minimized and the possible ill-conditioning is avoided for ma- trix inversion.Also,an effective weight algorithm with exponential function is proposed to reduce the time complexity of computing weight values and enhance its robustness for cross plane dataset. Comparative experiments show that WWLSTSVM obtains better resuits on eliminating the interference of noise samples and higher classification accuracy with less computing time in both linear and nonlinear cases compared with the other classifiers.