基于多模态深度学习的流量分类识别方法

A Traffic Classification and Recognition Method Based on Multimodal Deep Learning

针对流量分类效果与实际情况存在偏差的问题,首先将多模态深度学习运用在流量分类中,通过利用多模态之间的互补性,剔除模态间的冗余,从而学习到更好的流量数据特征表示。然后,提出了一种基于多模态流量数据的检测和分类方法,对同一流量单位的不同模态输入分别采用卷积神经网络(Convolutional Neural Networks,CNN)和长短期记忆网络(Long Short-Term Memory,LSTM)进行训练,以充分学习流量数据模态间和模态内信息的相互依赖性,克服现有单模态分类器的局限,从而支持更为复杂的现代网络应用场景。

aiming at the deviation of network traffic classification effect from the actual situation,we apply Multimodality Fusion Technology to learn a better feature representation of the traffic data firstly.Then we proposed a multimodal deep learning based traffic classifier using Convolutional Neutral Networks and Long Short-term Memory Recurrent Networks to learn both intra-and inter-modality dependences of traffic data.This approach can overcome performance limitations of existing single-modality deep learning based traffic classification proposals,and support more challenging traffic scenarios.