Analysis And Application Of User Mobility Based On Cellular Data Network Traffic

With the wide deployment of cellular data networks, we have witnessed the tremendous growth of mobile Internet access world.Meanwhile, there is great potential for network providers to capture big and invaluable data. Analyzing user mobility based on the cellular data network traffic has application potentials. However, it is challenging to have a deep understanding of user mobility based on cellular data network traffic. First, due to the loss of user identities and the phenomenon of "Data Island" between various online services, it is difficult to obtain complete online browsing behaviors and mobility information of users. Second, the datasets used by existing literature have the disadvantage of a limited pool of volunteers with similar living habits. Also, they have not discovered the spatio-temporal features of user mobility comprehensively and found out the distinct performances of different prediction model on various kinds of users yet. Third, there have been few in-depth investigations of the relationship between users' online content browsing behaviors and their real-life locations.Under this background, this dissertation first proposes a novel probability-based model to link virtual identities across online services in various service domains. Then, based on various kinds of contexts, the effects of contexts on users' spatio-temporal mobility prediction are analyzed. Further, this dissertation focuses on investigating the effect of living habits on the models of spatio-temporal prediction and next place prediction. Finally, we connect users' online behaviors with physical locations and give the first insight into the geospatial properties of online content browsing behaviors from the view of both geographic regions and users. The main research contents and innovations are as follows:(1) Design and implement users' virtual identity linkage system across service domains.In the era of the Internet, people are active in multiple online services,and usually have accounts on more than one online service. Each account is a virtual identity of the user. In order to trace individual's online behavior at any time and places, linking virtual identities belonging to the same natural person across different online service domains is very important.Existing methods usually tackle this problem by estimating the profile content similarity between identities under two different online services.However, the profile contents in various online services are unreliable or misaligned, and the proposed methods are always limited to services of a specific domain. In this thesis, we propose VISD, a method combining precise rules and probability-based rules, to link virtual identities across online services in various service domains. By using real-world network traffic data collected from two different kinds of networks, we evaluate the effectiveness of VISD model, which can achieve high linkage precision and recall.(2) Conduct spatio and temporal mobility analysis based on user mobile contexts.Mobile context is any information that can be used to characterize the situation of an individual, like current location or residence time in a certain place. Analyzing user mobility from the view of various mobile contexts helps us exploit the nature of human mobility and is essential to various mobile applications. This thesis first validates the effectiveness of cellular data network traffic on user mobility analysis. Then, we extract various spatial and temporal contexts, upon which multiple mobility predictors are established. Though the comparison of the performance between various models, we analyze the influence of spatial and temporal contexts on mobility prediction. Our finding shows that user mobility is highly related to both temporal and spatial contexts. Models considering the spatial context of current and previous locations can achieve high accuracy and robustness in both spatial and temporal prediction. Also, the spatial movement independent temporal predictors gain an edge for users who have a regular mobility pattern.(3) Predict individual future mobility at points of interest.A number of techniques have been proposed to either conduct spatio-temporal mobility prediction or forecast the next place. However, both of them produce diverse prediction performance for different users and display poor performance for some users. This thesis focuses on investigating the effect of living habits on the models of spatio-temporal prediction and next place prediction, and selects one from these two models for an individual to achieve effective mobility prediction at users' points of interest. Based on the hidden Markov model (HMM), a spatio-temporal predictor and a next place predictor are proposed. Living habits are analyzed in terms of entropy, upon which users are clustered into distinct groups. With large-scale factual mobile data captured from a big city, we compare the proposed HMM based predictors with existing state-of-the-art predictors, and apply them to different user groups. The results demonstrate the robust performance of the two proposed mobility predictors, which outperform the state of the art for various user groups.(4) Relate online content browsing behaviors to physical locations and measure its geospatial properties.With the growth of Mobile Internet, people are active in both the online and offline world. Investigating the relationships between users' online and offline behavior is a critical enabler for personalization, content caching,and urban planning. Although some studies have measured the spatial properties of online social relationships, the in-depth investigation of the relationship between online content browsing behavior and real-life location of users is limited. This paper gives the first insight into the geospatial properties of online content browsing behaviors from the point of view of both geographical regions and users. We first analyze the online browsing patterns across geographical regions and present a multilayer network based model to discover how inter-user distance affects the distribution of users with similar online browsing interests. Drawing upon results from comprehensive studies of users among three popular online content services in a metropolitan city of China, we achieve a comprehensive understanding of general and specific geospatial properties of users' various preferences: Users with similar online browsing interests have,to a large extent, strong geographic correlations, and the geospatial properties of different services are also distinct. The results of this work can be exploited to improve a potentially vast number of applications.