Cost-sensitive learning has been applied to resolve the multi-class imbalance problem in Internet traffic classification and it has achieved considerable results. But the classification performance on the minority cla...Cost-sensitive learning has been applied to resolve the multi-class imbalance problem in Internet traffic classification and it has achieved considerable results. But the classification performance on the minority classes with a few bytes is still unhopeful because the existing research only focuses on the classes with a large amount of bytes. Therefore, the class-dependent misclassification cost is studied. Firstly, the flow rate based cost matrix (FCM) is investigated. Secondly, a new cost matrix named weighted cost matrix (WCM) is proposed, which calculates a reasonable weight for each cost of FCM by regarding the data imbalance degree and classification accuracy of each class. It is able to further improve the classification performance on the difficult minority class (the class with more flows but worse classification accuracy). Experimental results on twelve real traffic datasets show that FCM and WCM obtain more than 92% flow g-mean and 80% byte g-mean on average; on the test set collected one year later, WCM outperforms FCM in terms of stability.展开更多
Rare categories become more and more abundant and their characterization has received little attention thus far. Fraudulent banking transactions, network intrusions, and rare diseases are examples of rare classes whos...Rare categories become more and more abundant and their characterization has received little attention thus far. Fraudulent banking transactions, network intrusions, and rare diseases are examples of rare classes whose detection and characterization are of high value. However, accurate char- acterization is challenging due to high-skewness and non- separability from majority classes, e.g., fraudulent transac- tions masquerade as legitimate ones. This paper proposes the RACH algorithm by exploring the compactness property of the rare categories. This algorithm is semi-supervised in na- ture since it uses both labeled and unlabeled data. It is based on an optimization framework which encloses the rare exam- ples by a minimum-radius hyperball. The framework is then converted into a convex optimization problem, which is in turn effectively solved in its dual form by the projected sub- gradient method. RACH can be naturally kernelized. Experi- mental results validate the effectiveness of RACH.展开更多
基金supported by the National Basic Research Program of China(2007CB307100,2007CB307106)
文摘Cost-sensitive learning has been applied to resolve the multi-class imbalance problem in Internet traffic classification and it has achieved considerable results. But the classification performance on the minority classes with a few bytes is still unhopeful because the existing research only focuses on the classes with a large amount of bytes. Therefore, the class-dependent misclassification cost is studied. Firstly, the flow rate based cost matrix (FCM) is investigated. Secondly, a new cost matrix named weighted cost matrix (WCM) is proposed, which calculates a reasonable weight for each cost of FCM by regarding the data imbalance degree and classification accuracy of each class. It is able to further improve the classification performance on the difficult minority class (the class with more flows but worse classification accuracy). Experimental results on twelve real traffic datasets show that FCM and WCM obtain more than 92% flow g-mean and 80% byte g-mean on average; on the test set collected one year later, WCM outperforms FCM in terms of stability.
文摘Rare categories become more and more abundant and their characterization has received little attention thus far. Fraudulent banking transactions, network intrusions, and rare diseases are examples of rare classes whose detection and characterization are of high value. However, accurate char- acterization is challenging due to high-skewness and non- separability from majority classes, e.g., fraudulent transac- tions masquerade as legitimate ones. This paper proposes the RACH algorithm by exploring the compactness property of the rare categories. This algorithm is semi-supervised in na- ture since it uses both labeled and unlabeled data. It is based on an optimization framework which encloses the rare exam- ples by a minimum-radius hyperball. The framework is then converted into a convex optimization problem, which is in turn effectively solved in its dual form by the projected sub- gradient method. RACH can be naturally kernelized. Experi- mental results validate the effectiveness of RACH.