动态代价支持向量机增量学习算法  

Incremental Support Vector Machine Algorithm on Dynamic Cost

作　　者：杜红乐[1] 张燕[1] 

机构地区：[1]商洛学院数学与计算机应用学院,陕西商洛726000

出　　处：《商洛学院学报》2017年第2期1-5,共5页Journal of Shangluo University

基　　金：陕西省教育厅专项科研计划项目(15JK1218)

摘　　要：在不均衡数据集下,SVM分类超平面的偏移,使得基于KKT条件进行样本选择的增量学习算法性能不佳,针对该问题,提出动态代价的SVM增量学习算法,该算法依据各类样本密度之间的关系动态计算类的错分代价,减少每次迭代中分类超平面偏移造成的错误累积,保证依据KKT条件选取样本的准确性,使得每次迭代选取的样本都包含当前分类器缺少的空间信息,提高最终分类器在不均衡数据集下的分类性能。最后,在UCI数据集上的仿真实验结果表明该算法能够提高不均衡数据下的分类性能。Under imbalanced dataset, SVM classification hyper plane shift in imbalanced data makes the poor performance of incremental learning algorithm based on KKT condition. As to this problem, an incremental support vector machine algorithm based on dynamic cost for imbalanced dataset is proposed. During each iteration, the algorithm dynamically calculates the class cost according to the density of samples and ensures the accuracy of selected samples with KKT conditions. It can improve the classification performance of the final classifier. Finally, experiment results with UCI dataset show the algorithm can improve the classification performance for imbalanced dataset.

关 键 词：支持向量机 不均衡数据集 增量学习 KKT条件 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]