论文部分内容阅读
针对当前自组织映射网络在流量分类中存在的不足,提出一种新的动态增长自组织映射模型DS-GSOM用于流量分类。该方法采用灵活可控的网络结构,引入调节因子EF来控制网络生长,可以按需要方便地在任意合适位置生成新结点,实现层次聚类,所生成的网络结点数目远远低于传统的SOFM方法,训练周期短,算法执行效率明显高于SOM和GSOM。实验分析结果表明该分类方法准确率和召回率明显优于自组织映射网络的其它流量分类方法。
Aiming at the shortcomings of current self-organizing mapping networks in traffic classification, a new dynamic growth self-organizing mapping model DS-GSOM is proposed for traffic classification. The method adopts a flexible and controllable network structure and introduces a regulation factor EF to control the growth of the network. The method can conveniently generate new nodes in any suitable position according to needs and realize hierarchical clustering. The generated network nodes have a much lower number of nodes than the traditional SOFM method, the training period is short, the execution efficiency of the algorithm is obviously higher than that of SOM and GSOM. Experimental results show that the accuracy and recall of this method are obviously better than those of other traffic classification methods in self-organizing mapping networks.