Search

With the popularity of smart mobile terminals and advances in wireless communication and positioning technologies, Geo-Social Networks (GSNs), which combine location awareness and social service functions, have become increasingly prevalent. The increasing amount of user and location information in GSNs makes the information overload phenomenon more and more serious. Although massive user-generated data brings convenience to users' social and travel activities, it also causes certain trouble for their daily life. In this context, users are expecting smarter mobile applications, so that the location information can be employed to perceive their surrounding environment intelligently and further mine their behavior patterns in GSNs, which ultimately provides personalized location-based services for users. Therefore, research on user location prediction comes into existence and has received extensive and in-depth attention from researchers. Through systematically analyzing the location data carried by user check-ins and comments, user location prediction can mine various user behavior patterns and personal preferences, thus determining the visiting location of users in the future. Research on user location prediction is still in the ascendant and it has become an important topic of common concern in both academia and industry. This survey takes Geo-social networking data as the focal point to elaborate the recent progress in user location prediction from multiple aspects such as problem categories, data sources, feature extraction, mathematical models and evaluation metrics. Besides, the difficulties to be studied and the future developmental trends of user location prediction are discussed. [ABSTRACT FROM AUTHOR]

Coreness is an important index to reflect the cohesiveness of a graph. The problems of core computation in static graphs and core update in dynamic graphs, known as the core decomposition and core maintenance problems respectively, have been extensively studied in previous work. However, most of these work focus on unweighted graphs. Considering that graphs are weighted in a lot of realistic applications, it is indispensable to extend the coreness to weighted graphs and devise efficient algorithms for weighted core decomposition and weighted core maintenance. In this work, we present a new definition of weighted coreness for vertices in a weighted graph, by taking into account the weights of vertices, which makes the coreness in unweighted graph be a special case. We propose efficient algorithms for both weighted core decomposition and weighted core maintenance problems. The coreness of vertices can be computed in linear time by the proposed decomposition algorithm, while the proposed core maintenance algorithm can process multiple-edge insertions/deletions simultaneously, which greatly reduces the core update time. Comprehensive experiments on both realistic networks and temporal graphs exhibit our algorithms are efficient and scalable. [ABSTRACT FROM AUTHOR]

Recently, the issue of offloading cellular data while reducing the duplicated cellular transmission has gained more and more attention. Several studies have shown that sharing contents through Device-to-Device (D2D) to offload traffic to local connections nearby can offer better performance for mobile users. Nevertheless, most existing proposals are somewhat confined to small-scale data sets or limited feature dimensions, relied on unconsolidated hypotheses and measurements of data sets. This paper presents a prior work of large-scale measurements on 3.56 TBytes of real-world data sets, which contain D2D content sharing activities from a popular D2D sharing application (APP). We conduct a comprehensive analysis of multi-dimensional features, including time series, structural properties, meeting dynamics, location relationship, and propagation tree. Our analysis reveals that (i) D2D sharing makes the hops between users shorter (in 3 or 4 degrees), (ii) the maximum spreading distance of content dissemination is 27 hops, (iii) we provide a new evidence of log-normal distribution of all user encounters (named meeting dynamics in this paper) based the fit of inter-TX time, Inter-Content Time (ICT) and Contact Time, (iv) online factor (O) and social factor (S) demonstrate the largest positive correlation and indicate that the two factors have high linear correlation. Finally, we analyze the correlations among all the impact factors by Pearson coefficient, principal component analysis, and latent semantic analysis, respectively. Results reveal that online factor (O) and social factor (S) have a high correlation, especially both of them have a great effect on D2D sharing activities. [ABSTRACT FROM AUTHOR]