【摘要】 针对传统的流量识别技术过于依赖个人的特征选择,无法同时满足实时性和准确性要求的问题,提出了一种基于残差神经网络（residual neural network,ResNet）模型的流量识别方法。根据网络流量数据和图像数据的相似性,对原始数据进行预处理,把一维的网络流量数据转换成二维的灰度图片,统一数据的输入格式;调整模型的超参数、训练模型的参数,筛选出最优的分类模型,实现对网络流量的识别。实验结果表明:该流量识别方法的准确率达到97. 03%,F_1-weighted值达到96. 54%,具有较高的识别率。通过与其他算法的结果对比,表明该方法的收敛速度快,识别准确率高,而且能够有效处理非均衡网络流量数据的识别问题。更多还原

【Abstract】 To solve the problem that traditional traffic recognition technology relies too much on individual feature selection and cannot meet the requirements of real-time and accuracy,a traffic recognition method based on Residual Neural Network( ResNet) model is proposed. According to the similarity between the network traffic data and the image data,the original data is preprocessed,and the one-dimensional network traffic data is converted into the two-dimensional gray image to unify the data input format. Adjust the parameters of the model and the training model to screen out the optimal classification model and realize the recognition of network traffic. The experimental results show that the accuracy of the method is 97. 03% and the value is 96. 54%,which has a high recognition rate. By comparing with the results of other algorithms,it is shown that this method has fast convergence speed,high recognition accuracy,and can effectively deal with the identification problem of unbalanced network traffic data.更多还原