基于多任务和卷积神经网络的业务识别算法  

Service recognition algorithm based on multi-task and convolutional neural network

作　　者：赵季红 乔琳琳[1] 王颖 ZHAO Jihong;QIAO Linlin;WANG Ying(School of Communications and Information Engineering,Xi'an University of Posts and Telecommunications,Xi'an 710121,China;School of Electronic and Information Engineering,Xi'an Jiaotong University,Xi'an 710049,China;Information Security Center of Beijing University of Posts and Telecommunications,Beijing 100876,China)

机构地区：[1]西安邮电大学通信与信息工程学院,陕西西安710121 [2]西安交通大学电子信息工程学院,陕西西安710049 [3]北京邮电大学信息安全中心,北京100876

出　　处：《西安邮电大学学报》2021年第1期1-6,共6页Journal of Xi’an University of Posts and Telecommunications

基　　金：国家自然科学基金项目(61531013);国家科技重大专项项目(2018ZX03001016)。

摘　　要：针对以低识别率对业务进行较高精度分类问题,提出了一种结合多任务学习和卷积神经网络(Multi-Task Learning and Convolutional Neural network,MTL-CNN)的网络业务识别算法,将业务分类重新构建为多任务学习框架,令业务类别作为主任务,带宽需求和持续时间作为辅助任务,3个任务在卷积神经网络(Convolutional Neural Network,CNN)中一起训练并进行预测,以此避免大量标记样本。仿真结果表明,所提算法对不同类别业务识别效果更加均衡,分类准确率达到95.60%。For the problem of classifying services with higher accuracy at low recognition rates,a network service recognition algorithm using multi-task learning and convolutional neural network(MTL-CNN)is proposed.By reconstructing the service classification as a multi-task learning framework,the proposed scheme formulates the business category as the main task,and builds the bandwidth requirements and duration as auxiliary tasks.Furthermore,the above three tasks are trained and predicted together in a convolutional neural network(CNN).This avoids the extra works of a large number of labeled samples.The simulation results show that the proposed algorithm has a more balanced effect on different types of business recognition,and the classification accuracy rate reaches 95.60%.

关 键 词：网络业务识别 多任务学习 卷积神经网络 低识别率 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]