基于总类内分布的松弛约束双支持向量机

Relaxed Constraint Twin Support Vector Machines Based on Total Within-class Distribution

针对数据分类问题,考虑到实际应用中噪声数据对分类结果的影响,提出一种新的基于总类内分布的松弛约束双支持向量机模型;该双支持向量机算法从约束不等式集出发,通过模糊集的思想引入一对约束参数项来松弛约束条件,提出松弛约束的隶属度函数,以有效减少噪声数据对分类结果的影响;同时将样本总的类内分布信息引入到双支持向量机模型的构造中,提出总类内离散度矩阵正定的条件。结果表明,与4个常见的双支持向量机相比,提出的双支持向量机模型不仅有较好的减噪及分类性能,而且具有较强的鲁棒性。

Aiming at the problem of data classification,considering the influence of noise data on classification results in practical applications,a new kind of relaxed constraint twin support vector machine model was proposed based on total within-class distribution. The twin support vector machine algorithm was based on the set of constraint inequalities. A pair of constraint parameter items was introduced by the theory of fuzzy sets,and the membership function of relaxation constraints was proposed to reduce the influence of noise data on classification resultseffectively. The total within-class distribution information was introduced into the construction of the twin support vector machine model,and the condition of the positive definite of the within-class scatter matrix was proposed. The results show that compared with the four common twin support vector machines,the proposed twin support vector machine model not only has better noise reduction and classification performance,but also has strong robustness.