Abstract | ||
---|---|---|
In view of inconsistent problems caused by that Synthetic Minority Over-sampling Technique (SMOTE) and Support Vector Machine (SVM) work in different space, this paper presents a kernel-based SMOTE approach to solve classification with imbalance data set by SVM. The method first preprocesses the data by oversampling the minority instances in the feature space, then the pre-images of the synthetic samples are found based on a distance relation between feature space and input space. Finally, these pre-images are appended to the original dataset to train a SVM. Experiments on real data set indicate that compared with SMOTE approach, the samples constructed by the proposed method have the higher quality. As a result, the effectiveness of classification by SVM on imbalance data set is improved. |
Year | DOI | Venue |
---|---|---|
2009 | 10.1007/978-3-642-10677-4_44 | ICONIP (1) |
Keywords | DocType | Volume |
support vector machine,input space,imbalance data,imbalance,imbalance data set,feature space,kernel-based smote approach,classification,pre-image.,improving svm classification,different space,smote approach,sampling technique | Conference | 5863 |
ISSN | Citations | PageRank |
0302-9743 | 16 | 0.63 |
References | Authors | |
10 | 2 |
Name | Order | Citations | PageRank |
---|---|---|---|
Zhiqiang Zeng | 1 | 139 | 16.35 |
Ji Gao | 2 | 50 | 9.03 |