Yue Liu

The rapid developments in mobile Internet are reshaping our lives and activities. Understanding the user behaviors and dynamics in such a large-scale network is essential for better system design, service provisioning, and network management. In this paper, we focus on the interaction pattern between mobile users and servers based on the traffic flow data. Real traffic flow data is collected from the public network of ISPs by high-performance network traffic monitors. Traffic Flow-based Heterogeneous Information Network (TF-HIN) is introduced to represent the traffic interaction pattern, and node correlation characteristics are mined from TF-HIN. Based on the empirical analysis of traffic interaction pattern, we propose the coupled flow tensor to represent the relations among user, server and time, by incorporating correlations of user and server as auxiliary information. Two iterative algorithms, i.e. FTD and FTD-NFS, are proposed for coupled flow tensor decomposition, and the latent factors are used for user clustering. We evaluate the proposed user clustering algorithms by using benchmark datasets, and also analyze the user clustering results from real traffic flow dataset. Numerical experiments show that the use of coupled flow tensor with auxiliary information provides a novel and scalable user clustering method and improves the clustering accuracy.

With the rapid development of wireless communication and mobile Internet, mobile phone becomes ubiquitous and functions as a versatile and smart system, on which people frequently interact with various mobile applications (Apps). Understanding human behaviors using mobile phone is significant for mobile system developers, for human-centered system optimization and better service provisioning. In this paper, we focus on mobile user behavior analysis and prediction based on mobile Internet traffic data. Real traffic flow data is collected from the public network of Internet Service Providers (ISPs), by high-performance network traffic monitors.We construct User-App bipartite network to represent the traffic interaction pattern between users and App servers. After mining the explicit and implicit features from User- App bipartite network, we propose two positive and unlabeled learning (PU learning) methods, including Spy-based PU learning and K-means-based PU learning, for App usage prediction and mobile video traffic identification. We firstly use the traffic flow data of QQ, a very famous messaging and social media application possessing high market share in China, as the experimental dataset for App usage prediction task. Then we use the traffic flow data from six popular Apps, including video intensive Apps (Youku, Baofeng, LeTV, Tudou) and other Apps (Meituan, Apple), as the experimental dataset for mobile video traffic identification task. Experimental results show that our proposed PU learning methods perform well in both tasks.