论文部分内容阅读
支持向量机(support vector machine,SVM)是分类算法中集高效性、准确率和实时性于一体的分类方案。但由于在SVM分类决策的过程中,无关的分类器也参与了投票,使得方案的实时性和分类可靠性有一定程度的降低。提出了基于相似度的高效SVM网络流量识别方案(efficient SVM based on similarity,ESVMS)。ESVMS通过估算待分类实例可能所属的类别范围,排除SVM中那些无关分类器的投票决策。实验结果表明ESVMS较SVM分类准确度几乎没有降低,但分类实时性进一步提高。
Support vector machine (SVM) is a classification scheme that integrates high efficiency, accuracy and real-time in the classification algorithm. However, in the process of SVM classification decision-making, irrelevant classifiers also participate in the voting, making the program’s real-time and classification reliability reduced to a certain extent. An efficient SVM based on similarity (ESVMS) algorithm based on similarity is proposed. ESVMS rules out voting decisions for those irrelevant classifiers in the SVM by estimating the range of categories to which the instances to be classified belong. The experimental results show that the accuracy of ESVMS is almost no lower than that of SVM classification, but the real-time classification is further improved.