计算机工程与设计
計算機工程與設計
계산궤공정여설계
COMPUTER ENGINEERING AND DESIGN
2014年
7期
2499-2503
,共5页
多类不平衡数据%支持向量机%空间扩展%小区快%上采样%SS-SVM算法
多類不平衡數據%支持嚮量機%空間擴展%小區快%上採樣%SS-SVM算法
다류불평형수거%지지향량궤%공간확전%소구쾌%상채양%SS-SVM산법
multi-class imbalance data%support vector machine%space-spreading%small-block%up-sampling%SS-SVM algorithm
针对多类不平衡数据分类准确率低的问题,提出一种基于空间扩展的支持向量机学习算法(support vector ma-chine algorithm based on space spreading,SS-SVM)。根据空间扩展原理,在多维欧式空间中通过空间扩展对少类数据进行上采样,使其处理数据时减少小区块的影响;降低数据不平衡度以优化分类器组;在扩展的数据集上训练SVM分类器。标准数据集上的实验结果表明,与几种经典的算法相比,SS-SVM在多类不平衡数据分类上可获得令人满意的分类结果,对少类数据分类精度要求较高的问题尤为有效。
針對多類不平衡數據分類準確率低的問題,提齣一種基于空間擴展的支持嚮量機學習算法(support vector ma-chine algorithm based on space spreading,SS-SVM)。根據空間擴展原理,在多維歐式空間中通過空間擴展對少類數據進行上採樣,使其處理數據時減少小區塊的影響;降低數據不平衡度以優化分類器組;在擴展的數據集上訓練SVM分類器。標準數據集上的實驗結果錶明,與幾種經典的算法相比,SS-SVM在多類不平衡數據分類上可穫得令人滿意的分類結果,對少類數據分類精度要求較高的問題尤為有效。
침대다류불평형수거분류준학솔저적문제,제출일충기우공간확전적지지향량궤학습산법(support vector ma-chine algorithm based on space spreading,SS-SVM)。근거공간확전원리,재다유구식공간중통과공간확전대소류수거진행상채양,사기처리수거시감소소구괴적영향;강저수거불평형도이우화분류기조;재확전적수거집상훈련SVM분류기。표준수거집상적실험결과표명,여궤충경전적산법상비,SS-SVM재다류불평형수거분류상가획득령인만의적분류결과,대소류수거분류정도요구교고적문제우위유효。
For the low accuracy of multi-class imbalanced data classification problem,a support vector machine algorithm was presented based on space spreading,namely SS-SVM.Through space spreading in multi-dimensional Euclidean space based on space spreading principle,SS-SVM algorithm added the size of small class dataset by using up-sampling method to reduce the im-pact of the small-block,and the imbalance of datasets was reduced.Then the classifier group was optimized and the SVM classi-fier was trained on this solved dataset.The experiment results on UCI datasets demonstrated that the SS-SVM had satisfactory classification performance comparing with several traditional algorithms.Especially,it was more efficient for some data proces-sing problems,for example,the predicting precision of the small class dataset should not be under a threshold.