基于三支决策的支持向量机增量学习方法  

Three-way Decisions-based Incremental Learning Method for Support Vector Machine

作　　者：徐久成 刘洋洋[1,2] 杜丽娜 孙林[1,2] 

机构地区：[1]河南师范大学计算机与信息工程学院,新乡453007 [2]河南省高校计算智能与数据挖掘工程技术研究中心,新乡453007

出　　处：《计算机科学》2015年第6期82-87,共6页Computer Science

基　　金：国家自然科学基金项目(61370169;61402153;60873104);河南省科技攻关重点项目(142102210056);新乡市重点科技攻关计划项目(ZG13004)资助

摘　　要：针对典型的支持向量机增量学习算法对有用信息的丢失和现有支持向量机增量学习算法单纯追求分类器精准性的客观性,将三支决策损失函数的主观性引入支持向量机增量学习算法中,提出了一种基于三支决策的支持向量机增量学习方法。首先采用特征距离与中心距离的比值来计算三支决策中的条件概率;然后把三支决策中的边界域作为边界向量加入到原支持向量和新增样本中一起训练;最后,通过仿真实验证明,该方法不仅充分利用有用信息提高了分类准确性,而且在一定程度上修正了现有支持向量机增量学习算法的客观性,并解决了三支决策中条件概率的计算问题。Aiming at the problems that the typical incremental learning algorithm for support vector machine （SVM） loses a lot of useful information and the objectivity that existing incremental learning algorithms for SVM aspire to clas- sification accuracy merely, the subjectivity of loss functions of three-way decisions was introduced to incremental lear- ning algorithms for SVM,and a three-way decisions-based incremental learning method for SVM was proposed. Firstly the conditional probability of three-way decisions was denoted by the ratio of feature distances and center distances. Sec- ondly the objects of boundary region of three-way decisions were regarded as boundary vectors to be trained with the original support vectors and the newly added samples. Finally, simulation experiments were done. The results show that the proposed method not only makes full use of the useful information to improve the classification accuracy, but also revises the objectivity of existing incremental learning algorithms for SVM to some extent. Besides, the computation problem of conditional probability of three-way decisions is resolved.

关 键 词：三支决策 支持向量机 增量学习 条件概率 边界向量 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]