基于链式结构的网络视频流分类算法

Internet video traffic classification algorithm based on chain structure

针对各类别网络流分布不平衡的问题,设计了一种能够实现低存储、低时延、高准确率的网络视频流细分类算法.首先,采用改进的卡方离散算法对数据进行离散化处理;然后,提出了一种改进线性前向特征选择算法,选出有效的QoS相关特征;最后,设计一种链式和分级结构相结合的分类结构,完成网络视频流细分类.针对真实网络采集的7种网络视频流的分类试验结果表明,所提算法的分类准确率达到96.7%,而且在数据不平衡的情况下仍具有较高的识别率.

As for the imbalanced distribution of each category of internet flows, an internet video traffic classification algorithm with small memory, low latency and high classification accuracy was designed. First, a modified Chi2 discretization algorithm was used to discretize data. Then, a modified linear forward selection(MLFS) method was proposed to select effective QoS(quality of service) features. Finally, a classification structure combing a chain structure with a hierarchical structure was designed to achieve fine-grained classification of internet video flows. The experimental results of the traffic classification with seven categories in real-network dataset show that the classification accuracy of the proposed algorithm is 96.7%. And the high recognition rate can keep stable for the imbalanced dataset.