Mobile traffic analysis and measurement

Mobile traffic analysis and measurement

Mobile market is driven to be explosive by the prosperity of hand-held smart devices and full-experience mobile applications. Traditional problems such as networking optimization and human social analysis can be tackled in new ways. From an abstract view, users with mobile devices are sophisticated sensors of our realistic world. The convenience of accessing human-generated data via mobile equipment stimulates rapid development of ubiquitous computing technologies and human behaviour researches. For example, the success of revenue models based on subscription and advertisement relies heavily upon the knowledge of user participation and individual preferences. In mobile networks, accurate interpretation of user behaviour and experience is of use to facilitate network management for Internet service providers (ISPs), application decision optimization for content providers, and the enhancement of user privacy protection policies from a hacking view.

Since 2011, our researchers in OMNILab and academic partners have made extensive steps towards understanding UX and spatiotemporal characteristics of mobile traffic. Recently, we proposed the concept of mobile user engagement in designing human-concentrated mobile technologies and applications. We focused on a characterization and modelling to understand mobile user engagement, especially its interactions with user-perceived application performance (a.k.a. user perception) in real context. At the urban scale, limited knowledge about the spatiotemporal dependence of mobile traffic is gained. Our researchers fill this gap by making use of request-response records extracted from HTTP traffic at the city scale. The results suggest connections between spatio- temporal dependence of cellular traffic and the organization of human lives. Region differences are observed to impact traffic dependence to a great extent. Additionally, interactive knowledge between space and time enhances traffic prediction with a decrease in root-mean-square error of 2.8%~25.2%.

随着移动技术的快速发展,传统的网络优化和人类行为研究,可以借助移动技术产生的海量数据从一个新的视角进行研究。从抽象角度来看,移动网络的用户 如同一个个现实世界的传感器,数字化这我们生活中各种各样的行为。这种数据获取的便利,促进了普适计算技术(ubiquitous computing)和人类行为学的快速发展。例如,一个传统意义上基于广告推广营销的平台,对用户参与行为和个人爱好有着很强的依赖。基于移动数据的分 析便可以较小的代价获得更大的收益。在这样的应用中,对用户行为的准确解读是这类技术成功与否的关键。与此同时,这类技术对服务运营商和内容分发服务商、 甚至用户隐私的保护策略,都有着重要的价值。

自2011年以来,OMNILab的研究人员以及合作伙伴,在移动网络流量分析和用户体验测量上进行了深入的研究。我们利用移动用户参与度 (user engagement)的概念,从网络流量出发对用户角度的体验进行客观测量和分析。在我们的方法框架中,用户参与行为、网络性能以及使用场景得到了细致 的分类研究以及综合模型分析,从相关性角度找出了影响用户参与度的关键影响因素。除此以外,大范围角度的移动流量特征也是OMNILab研究人员关注的领 域。受限于传统网络上数据采集规模的限制,城市角度的流量时空特征一直未能得到充分地理解。我们基于中国某城市的移动网络流量数据,对网络流量的时间-空 间依赖性(spatiotemporal dependence)进行了系统分析。结果显示,大尺度上的移动流量分布是非随机和非线性的(如幂律分布),而且和人类的日常行为紧密联系在一起。在结 合了时空依赖信息以后,我们对移动网络流量的预测均方误差减小了2.8%~25.2%。移动网络流量分析是我们的长期研究领域,希望能和更多的合作伙伴共 享我们的技术,将研究成果用于实际当中

Person/Organization: Xiaming Chen, Siwei Qiang, Xin Yang
Project:

  • Mobile traffic mining funded by Huawei Inc.
  • HTTP traffic analysis funded by China Telecom

Papers:

  • Chen, Xiaming; Jin, Yaohui; Qiang, Siwei; Hu, Weisheng; Jiang, Kaida; Analyzing and Modeling Spatio-Temporal Dependence of Cellular Traffic at City Scale, Submission to ICC 2015.
  • Chen, Xiaming; Qiang, Siwei; Wei, Jianwen; Jiang, Kaida; Jin, Yaohui; Towards Understanding User Engagement in Mobile Networks: A Perception Perspective, Submission to Elsevier Pervasive and Mobile Computing.
  • Chen, Xiaming; Jin, Yaohui; Web Page Identification from Network Traffic with Support Vector Machine,Technical Report, 2012, Shanghai Jiao Tong Univ.
  • Yang, Xin; Chen, Xiaming; Jin, Yaohui; A high-speed real-time HTTP performance measurement architecture based on network processor, ICT Convergence (ICTC), 2011 International Conference on,744-745,2011,IEEE.