Author(s): Yuan Yao | Lin Feng | Bo Jin | Feng Chen
Journal: Information Technology Journal
ISSN 1812-5638
Volume: 11;
Issue: 2;
Start page: 200;
Date: 2012;
VIEW PDF
DOWNLOAD PDF 
Original page
Keywords: support vector machine | data stream | multiple classifiers model | concept drift | Incremental learning | network intrusion detection
ABSTRACT
Currently, data mining for data streams has gained importance in the network management area. Although many new technologies have been applied in this area, most of them belong to the rule-based style. In order to overcome the weakness of rule-based mode, the learning-based model with incremental learning method was employed. In this study, the model proposed was optimized in Support Vector Machine (SVM) kernel functions selection and the parameters. Apart from this, real world network data sets were used in the experiment to certify the validity of the new model. The experimental result showed that the optimized model can improve the accuracy of classification and reduce the time cost. At the same time, the optimized model was also compared with other models.
Journal: Information Technology Journal
ISSN 1812-5638
Volume: 11;
Issue: 2;
Start page: 200;
Date: 2012;
VIEW PDF
DOWNLOAD PDF Original pageKeywords: support vector machine | data stream | multiple classifiers model | concept drift | Incremental learning | network intrusion detection
ABSTRACT
Currently, data mining for data streams has gained importance in the network management area. Although many new technologies have been applied in this area, most of them belong to the rule-based style. In order to overcome the weakness of rule-based mode, the learning-based model with incremental learning method was employed. In this study, the model proposed was optimized in Support Vector Machine (SVM) kernel functions selection and the parameters. Apart from this, real world network data sets were used in the experiment to certify the validity of the new model. The experimental result showed that the optimized model can improve the accuracy of classification and reduce the time cost. At the same time, the optimized model was also compared with other models.
