Citation Context

...information. Precision-recall. Fig. 14 summarizes the precision-recall results for users with community entropies higher than 1.0. The posterior probability of SVM is calculated by the Platt’s method =-=[23]-=-. First, our location predictor based on community data performs better than the one based on friends’ information in the four metropolises. Second, the communitybased model’s performance itself is pr...

Citation Context

...: they all treat friends of users equally. First, in online social networks some of a user’s friends are not his real friends, and the links may be created randomly or due to other malicious purposes =-=[17]-=-. These friends have little influence on users’ behaviors. Second, even only considering real friends, similar to other social behaviors, in most cases mobility is only influenced by communities but n...

Citation Context

...characterize users’ social diversity in Section 3.2. 3.1 Community detection in social networks Community detection in networks (or graphs) has been extensively studied for the past decade (e.g., see =-=[25, 31, 2, 18, 32, 22, 38, 37, 21, 39, 23]-=-). It has important applications in many fields, including physics, biology, sociology as well as computer science. The principle behind community detection is to partition nodes of a large graph into...

Citation Context

...characterize users’ social diversity in Section 3.2. 3.1 Community detection in social networks Community detection in networks (or graphs) has been extensively studied for the past decade (e.g., see =-=[25, 31, 2, 18, 32, 22, 38, 37, 21, 39, 23]-=-). It has important applications in many fields, including physics, biology, sociology as well as computer science. The principle behind community detection is to partition nodes of a large graph into...

Citation Context

...ursquare. Since these large amount of location and social relation data become available, studying human mobility and its connection with social relationships becomes quantitatively achievable (e.g., =-=[16, 10, 9, 34, 15, 7, 8]-=-). Understanding human mobility can lead to compelling applications including location recommendation [44, 41, 43, 14, 20], urban planning [42], immigration patterns [5], etc. Previous works, includin...

Citation Context

...e (e.g., [16, 10, 9, 34, 15, 7, 8]). Understanding human mobility can lead to compelling applications including location recommendation [44, 41, 43, 14, 20], urban planning [42], immigration patterns =-=[5]-=-, etc. Previous works, including [1, 6, 10, 33], show that human mobility is influenced by social factors. However, there is one common shortcoming: they all treat friends of users equally. Similar to...

Citation Context

...ursquare. Since these large amount of location and social relation data become available, studying human mobility and its connection with social relationships becomes quantitatively achievable (e.g., =-=[16, 10, 9, 34, 15, 7, 8]-=-). Understanding human mobility can lead to compelling applications including location recommendation [44, 41, 43, 14, 20], urban planning [42], immigration patterns [5], etc. Previous works, includin...

Citation Context

...characterize users’ social diversity in Section 3.2. 3.1 Community detection in social networks Community detection in networks (or graphs) has been extensively studied for the past decade (e.g., see =-=[25, 31, 2, 18, 32, 22, 38, 37, 21, 39, 23]-=-). It has important applications in many fields, including physics, biology, sociology as well as computer science. The principle behind community detection is to partition nodes of a large graph into...

Citation Context

...ction with social relationships becomes quantitatively achievable (e.g., [16, 10, 9, 34, 15, 7, 8]). Understanding human mobility can lead to compelling applications including location recommendation =-=[44, 41, 43, 14, 20]-=-, urban planning [42], immigration patterns [5], etc. Previous works, including [1, 6, 10, 33], show that human mobility is influenced by social factors. However, there is one common shortcoming: they...

Citation Context

... Understanding human mobility can lead to compelling applications including location recommendation [44, 41, 43, 14, 20], urban planning [42], immigration patterns [5], etc. Previous works, including =-=[1, 6, 10, 33]-=-, show that human mobility is influenced by social factors. However, there is one common shortcoming: they all treat friends of users equally. Similar to other social behaviors, in most cases mobility...

Citation Context

...ve been intensively studied [9, 34, 15]. There are mainly two directions of research going on in the area. One direction is to use the location information from LBSNs to predict friendships (see e.g. =-=[19, 13, 12, 6, 33, 28, 40]-=-), the other studies the impact from friendships on locations [1, 6, 10, 33, 24] which is what we focus on in the current work. Backstrom, Sun and Marlow [1] study the friendship and location using th...

Citation Context

...ction with social relationships becomes quantitatively achievable (e.g., [16, 10, 9, 34, 15, 7, 8]). Understanding human mobility can lead to compelling applications including location recommendation =-=[44, 41, 43, 14, 20]-=-, urban planning [42], immigration patterns [5], etc. Previous works, including [1, 6, 10, 33], show that human mobility is influenced by social factors. However, there is one common shortcoming: they...

Citation Context

...ursquare. Since these large amount of location and social relation data become available, studying human mobility and its connection with social relationships becomes quantitatively achievable (e.g., =-=[16, 10, 9, 34, 15, 7, 8]-=-). Understanding human mobility can lead to compelling applications including location recommendation [44, 41, 43, 14, 20], urban planning [42], immigration patterns [5], etc. Previous works, includin...

Citation Context

...characterize users’ social diversity in Section 3.2. 3.1 Community detection in social networks Community detection in networks (or graphs) has been extensively studied for the past decade (e.g., see =-=[25, 31, 2, 18, 32, 22, 38, 37, 21, 39, 23]-=-). It has important applications in many fields, including physics, biology, sociology as well as computer science. The principle behind community detection is to partition nodes of a large graph into...

Citation Context

...ve been intensively studied [9, 34, 15]. There are mainly two directions of research going on in the area. One direction is to use the location information from LBSNs to predict friendships (see e.g. =-=[19, 13, 12, 6, 33, 28, 40]-=-), the other studies the impact from friendships on locations [1, 6, 10, 33, 24] which is what we focus on in the current work. Backstrom, Sun and Marlow [1] study the friendship and location using th...

Citation Context

...and social relation data become available, studying human mobility, one of the most common behaviors of human being, and its connection with social relationships become quantitatively possible (e.g., =-=[13, 7, 6, 26, 11, 5]-=-). Understanding human mobility can lead to compelling applications including location recommendation [34, 31, 33, 10, 16], urban planning [32], immigration patterns [3], etc. Previous works, includin...

Citation Context

...contexts. 5. LOCATION PREDICTION Location prediction can drive compelling applications including location recommendation and targeted advertising. On the other hand, it may also threat users’ privacy =-=[35]-=-. Following the previous analysis, we continue to investigate whether it is possible to use community information to effectively predict users’ locations, using machine learning techniques. More preci...

Citation Context

...ve been intensively studied [9, 34, 15]. There are mainly two directions of research going on in the area. One direction is to use the location information from LBSNs to predict friendships (see e.g. =-=[19, 13, 12, 6, 33, 28, 40]-=-), the other studies the impact from friendships on locations [1, 6, 10, 33, 24] which is what we focus on in the current work. Backstrom, Sun and Marlow [1] study the friendship and location using th...

Citation Context

...ction with social relationships becomes quantitatively achievable (e.g., [16, 10, 9, 34, 15, 7, 8]). Understanding human mobility can lead to compelling applications including location recommendation =-=[44, 41, 43, 14, 20]-=-, urban planning [42], immigration patterns [5], etc. Previous works, including [1, 6, 10, 33], show that human mobility is influenced by social factors. However, there is one common shortcoming: they...

Citation Context

... Understanding human mobility can lead to compelling applications including location recommendation [44, 41, 43, 14, 20], urban planning [42], immigration patterns [5], etc. Previous works, including =-=[1, 6, 10, 33]-=-, show that human mobility is influenced by social factors. However, there is one common shortcoming: they all treat friends of users equally. Similar to other social behaviors, in most cases mobility...

Citation Context

...ion flow among communities and within each community is minimized. Since it is infeasible to search all possible community partitions, Infomap further exploits a deterministic greedy search algorithm =-=[11, 36]-=- to find partitions. In our work, to detect communities of u, we first find all his friends as well as the links among them. Then, we delete u and all edges linked to him and apply Infomap algorithm t...

Citation Context

...here are mainly two reasons for the existence of fake friends. First, some users treat online links as a capital and they try to add as many friends as possible to become powerful in the online world =-=[12]-=-. Second, some users intend to threat some target’s privacy by adding his friends into their social networks [17]. For a normal user, the above mentioned users linked to him are called his fake friend...

Citation Context

...characterize users’ social diversity in Section 3.2. 3.1 Community detection in social networks Community detection in networks (or graphs) has been extensively studied for the past decade (e.g., see =-=[25, 31, 2, 18, 32, 22, 38, 37, 21, 39, 23]-=-). It has important applications in many fields, including physics, biology, sociology as well as computer science. The principle behind community detection is to partition nodes of a large graph into...

Citation Context

...ursquare. Since these large amount of location and social relation data become available, studying human mobility and its connection with social relationships becomes quantitatively achievable (e.g., =-=[16, 10, 9, 34, 15, 7, 8]-=-). Understanding human mobility can lead to compelling applications including location recommendation [44, 41, 43, 14, 20], urban planning [42], immigration patterns [5], etc. Previous works, includin...

Citation Context

...characterize users’ social diversity in Section 3.2. 3.1 Community detection in social networks Community detection in networks (or graphs) has been extensively studied for the past decade (e.g., see =-=[25, 31, 2, 18, 32, 22, 38, 37, 21, 39, 23]-=-). It has important applications in many fields, including physics, biology, sociology as well as computer science. The principle behind community detection is to partition nodes of a large graph into...

Citation Context

... Understanding human mobility can lead to compelling applications including location recommendation [44, 41, 43, 14, 20], urban planning [42], immigration patterns [5], etc. Previous works, including =-=[1, 6, 10, 33]-=-, show that human mobility is influenced by social factors. However, there is one common shortcoming: they all treat friends of users equally. Similar to other social behaviors, in most cases mobility...

Citation Context

...ction with social relationships becomes quantitatively achievable (e.g., [16, 10, 9, 34, 15, 7, 8]). Understanding human mobility can lead to compelling applications including location recommendation =-=[44, 41, 43, 14, 20]-=-, urban planning [42], immigration patterns [5], etc. Previous works, including [1, 6, 10, 33], show that human mobility is influenced by social factors. However, there is one common shortcoming: they...

Citation Context

...characterize users’ social diversity in Section 3.2. 3.1 Community detection in social networks Community detection in networks (or graphs) has been extensively studied for the past decade (e.g., see =-=[25, 31, 2, 18, 32, 22, 38, 37, 21, 39, 23]-=-). It has important applications in many fields, including physics, biology, sociology as well as computer science. The principle behind community detection is to partition nodes of a large graph into...

Citation Context

...characterize users’ social diversity in Section 3.2. 3.1 Community detection in social networks Community detection in networks (or graphs) has been extensively studied for the past decade (e.g., see =-=[25, 31, 2, 18, 32, 22, 38, 37, 21, 39, 23]-=-). It has important applications in many fields, including physics, biology, sociology as well as computer science. The principle behind community detection is to partition nodes of a large graph into...

Citation Context

...es quantitatively achievable (e.g., [16, 10, 9, 34, 15, 7, 8]). Understanding human mobility can lead to compelling applications including location recommendation [44, 41, 43, 14, 20], urban planning =-=[42]-=-, immigration patterns [5], etc. Previous works, including [1, 6, 10, 33], show that human mobility is influenced by social factors. However, there is one common shortcoming: they all treat friends of...

Citation Context

...heir observation from the geo-tagged Twitter data and incorporate it into the model to predict users’ home locations. Experimental results show that their model outperforms the one of [1]. Jurgens in =-=[17]-=- proposes a spatial label propagation algorithm to infer a user’s location based on a small number initial friends’ locations. Techniques such as exploiting information from multiple social network pl...

Citation Context

...ursquare. Since these large amount of location and social relation data become available, studying human mobility and its connection with social relationships becomes quantitatively achievable (e.g., =-=[16, 10, 9, 34, 15, 7, 8]-=-). Understanding human mobility can lead to compelling applications including location recommendation [44, 41, 43, 14, 20], urban planning [42], immigration patterns [5], etc. Previous works, includin...

Citation Context

...ve been intensively studied [9, 34, 15]. There are mainly two directions of research going on in the area. One direction is to use the location information from LBSNs to predict friendships (see e.g. =-=[19, 13, 12, 6, 33, 28, 40]-=-), the other studies the impact from friendships on locations [1, 6, 10, 33, 24] which is what we focus on in the current work. Backstrom, Sun and Marlow [1] study the friendship and location using th...

Citation Context

...ction with social relationships becomes quantitatively achievable (e.g., [16, 10, 9, 34, 15, 7, 8]). Understanding human mobility can lead to compelling applications including location recommendation =-=[44, 41, 43, 14, 20]-=-, urban planning [42], immigration patterns [5], etc. Previous works, including [1, 6, 10, 33], show that human mobility is influenced by social factors. However, there is one common shortcoming: they...

Citation Context

...g on in the area. One direction is to use the location information from LBSNs to predict friendships (see e.g. [19, 13, 12, 6, 33, 28, 40]), the other studies the impact from friendships on locations =-=[1, 6, 10, 33, 24]-=- which is what we focus on in the current work. Backstrom, Sun and Marlow [1] study the friendship and location using the Facebook data with user-specified home addresses. They find out that the frien...

Citation Context

...certain locations in the future not their home [1, 24, 17] or a dynamic sequences of locations [33]. We focus on understanding users’ mobility behavior from social network communities. The authors of =-=[4]-=- tackle the inverse problem, i.e., they exploit users’ mobility information to detect communities. They first attach weights to the edges in a social network based on the check-in information, then th...

Citation Context

...ding to the comparative analysis [18], among all the community detection algorithms, Infomap [31] has the best performance on undirected and unweighted graphs and has been widely used in many systems =-=[26, 29]-=-. Therefore, we apply it in this work. Next we give a brief overview of Infomap and describe how we use it to detect communities. The main idea of Infomap can be summarized as follows: information flo...

Citation Context

...act on a user’s mobility should be considered from the perspectives of communities instead of all friends. In a broader view, community is arguably the most useful resolution to study social networks =-=[39]-=-. Contributions. In this paper, we aim to study the impact from communities on a user’s mobility and predict his locations based on his community information. First, we partition each users’ friends i...

Citation Context

...ve been intensively studied [9, 34, 15]. There are mainly two directions of research going on in the area. One direction is to use the location information from LBSNs to predict friendships (see e.g. =-=[19, 13, 12, 6, 33, 28, 40]-=-), the other studies the impact from friendships on locations [1, 6, 10, 33, 24] which is what we focus on in the current work. Backstrom, Sun and Marlow [1] study the friendship and location using th...

Citation Context

... results show that their method is able to discover more meaningful communities, such as place-focused communities, compared to the standard community detection algorithm. More recently, Brown et al. =-=[3]-=- analyze mobility behaviors of pairs of friends and groups of friends (communities). They focus on comparing the difference between individual mobility and group mobility. For example, they discover t...

Citation Context

...characterize users’ social diversity in Section 3.2. 3.1 Community detection in social networks Community detection in networks (or graphs) has been extensively studied for the past decade (e.g., see =-=[25, 31, 2, 18, 32, 22, 38, 37, 21, 39, 23]-=-). It has important applications in many fields, including physics, biology, sociology as well as computer science. The principle behind community detection is to partition nodes of a large graph into...

Citation Context

...ding to the comparative analysis [18], among all the community detection algorithms, Infomap [31] has the best performance on undirected and unweighted graphs and has been widely used in many systems =-=[26, 29]-=-. Therefore, we apply it in this work. Next we give a brief overview of Infomap and describe how we use it to detect communities. The main idea of Infomap can be summarized as follows: information flo...

Citation Context

... of community entropy. Definition 1. For a user u, his community entropy is defined as coment(u) = 1 1− α ln ∑ c∈C(u) ( |c| |f (u)| ) α. Our community entropy follows the definition of Rényi entropy =-=[30]-=-. Here, α is called the order of diversity, it can control the impact of community size on the value which gives more flexibility to distinguish users when focusing on the sizes of their communities. ...