基于ν-最大间隔超球体支持向量机的非平衡数据分类
|
摘要
针对传统超球体支持向量机在处理非平衡数据集时出现的少数类查全率较低的问题,通过引入最大间隔和参数ν,提出了ν-最大间隔超球体支持向量机。该算法以最大间隔为优化目标建立分类模型,大幅度提高了少数类查全率。通过UCI数据集的分类仿真实验,比较了该算法与传统超球体支持向量机的分类精度。结果表明,该算法能有效提高对非均衡分布数据的分类准确率。
In view of the traditional HSSVM(hyper sphere support vector machine) in processing of imbalanced data set the minority class the problem of low recall ratio,and by introducing the maximum margin and the parameter nu,this paper proposed the NU-MMHSSVM(nu-maximum margin hyper sphere support vector machine).The algorithm takes the maximum margin as the optimization target to build a classification model,greatly improving the minority class recall.Through UCI data set classification experiments,the algorithm with the traditional HSSVM(hyper sphere support vector machine) classification accuracy is compared.The results show that the algorithm can effectively improve the imbalanced distribution of data classification accuracy.
In view of the traditional HSSVM(hyper sphere support vector machine) in processing of imbalanced data set the minority class the problem of low recall ratio,and by introducing the maximum margin and the parameter nu,this paper proposed the NU-MMHSSVM(nu-maximum margin hyper sphere support vector machine).The algorithm takes the maximum margin as the optimization target to build a classification model,greatly improving the minority class recall.Through UCI data set classification experiments,the algorithm with the traditional HSSVM(hyper sphere support vector machine) classification accuracy is compared.The results show that the algorithm can effectively improve the imbalanced distribution of data classification accuracy.
引文
[1]Vapnik V N.The Nature of Statistical Learning Theory[M].London,UK:Springer-Verlag,1995.
[2]邬啸,魏延,吴瑕.基于混合核函数的支持向量机[J].重庆理工大学学报:自然科学版,2011(10):66-70.
[3]余珺,郑先斌,张小海.基于多核优选的装备费用支持向量机预测法[J].四川兵工学报,2011(6):118-119.
[4]万辉.一种基于最小二乘支持向量机的图像增强算法[J].重庆理工大学学报:自然科学版,2011(6):53-57.
[5]罗沛清,梁青阳,江钦龙,等.基于分层聚类的支持向量机模拟电路故障诊断[J].四川兵工学报,2011(9):92-95..
[6]崔建国,李明,陈希成.基于支持向量机的飞行器健康诊断方法[J].压电与声光,2009(2):266-269.
[7]张宏蕾,张立亭,罗亦泳,等.基于支持向量机的土地利用预警研究[J].安徽农业科学,2010(35):20503-20504.
[8]唐晓芬,赵秉新.基于支持向量机的农村劳动力转移预测[J].安徽农业科学,2011(11):6837-6838.
[9]Tax D,Duin R.Support vector domain description[J].Pattern Recognition Letters,2003,20:11-13.
[10]Tax D,Duin R.Support vector domain description[J].Machine Leaning,2004(1):45-66.
[11]文传军,詹永照,陈长军.最大间隔最小体积球形支持向量机[J].控制与决策,2010,25(1):79-83.
[12]张慧敏,柴毅.不等距超球体支持向量机[J].计算机工程与应用,2011,47(11):19-22.
[13]王娟,胡文军,王士同.最大边界模糊核超球分类方法[J].计算机应用2011,31(9):2542-2545.
[14]Collobert R,Bengio S.SVMTorch:Support vector machinefor large-scale regression problems[J].J of MachineLearning Research,2001,1(2):143-160.
[15]Frank A.Asuncion A UCI repository of machine learningdatabases[EB/OL].[2012-06-18].http://archive.ics.uci.edu/ml.
[16]Joshi M V.On Evaluating Performance of Classifiers forRare Classes[C]//Proc of the 2nd IEEE InternationalConference on Data Mining.Maebishi,Japan:[s.n.],2002:641-644.
[2]邬啸,魏延,吴瑕.基于混合核函数的支持向量机[J].重庆理工大学学报:自然科学版,2011(10):66-70.
[3]余珺,郑先斌,张小海.基于多核优选的装备费用支持向量机预测法[J].四川兵工学报,2011(6):118-119.
[4]万辉.一种基于最小二乘支持向量机的图像增强算法[J].重庆理工大学学报:自然科学版,2011(6):53-57.
[5]罗沛清,梁青阳,江钦龙,等.基于分层聚类的支持向量机模拟电路故障诊断[J].四川兵工学报,2011(9):92-95..
[6]崔建国,李明,陈希成.基于支持向量机的飞行器健康诊断方法[J].压电与声光,2009(2):266-269.
[7]张宏蕾,张立亭,罗亦泳,等.基于支持向量机的土地利用预警研究[J].安徽农业科学,2010(35):20503-20504.
[8]唐晓芬,赵秉新.基于支持向量机的农村劳动力转移预测[J].安徽农业科学,2011(11):6837-6838.
[9]Tax D,Duin R.Support vector domain description[J].Pattern Recognition Letters,2003,20:11-13.
[10]Tax D,Duin R.Support vector domain description[J].Machine Leaning,2004(1):45-66.
[11]文传军,詹永照,陈长军.最大间隔最小体积球形支持向量机[J].控制与决策,2010,25(1):79-83.
[12]张慧敏,柴毅.不等距超球体支持向量机[J].计算机工程与应用,2011,47(11):19-22.
[13]王娟,胡文军,王士同.最大边界模糊核超球分类方法[J].计算机应用2011,31(9):2542-2545.
[14]Collobert R,Bengio S.SVMTorch:Support vector machinefor large-scale regression problems[J].J of MachineLearning Research,2001,1(2):143-160.
[15]Frank A.Asuncion A UCI repository of machine learningdatabases[EB/OL].[2012-06-18].http://archive.ics.uci.edu/ml.
[16]Joshi M V.On Evaluating Performance of Classifiers forRare Classes[C]//Proc of the 2nd IEEE InternationalConference on Data Mining.Maebishi,Japan:[s.n.],2002:641-644.