Accurately predicting traffic conditions is vital for smart city development, yet it remains challenging due to the intricate spatio-temporal dependencies in road networks. Existing works often propose intra-mixing deep learning-based prediction models for individual nodes and share parameters among them or spatial intermixing deep learning-based models for traffic predictions. However, these approaches may neglect essential principles of information exchange in traffic flow or capture useless or even erroneous spatio-temporal dependencies. To address these limitations, we propose a Mixing Spatio-Temporal neural network (MiST) for enhancing traffic predictions. In MiST, we propose (i) a temporal encoder that embeds the traffic data along with periodic features, (ii) a spatial encoder that embeds the positional information in graph and hypergraph spectral domains, as well as spatial node identities, and (iii) a mixing spatio-temporal encoder that merges the diverse features provided by the temporal and spatial encoders. Our empirical evaluations on real-world traffic prediction tasks, including flow and speed predictions, validate the superiority of MiST, underscoring its innovative contribution to traffic prediction methodologies.

Xiancai Xu, Jia-Dong Zhang, Lei Xiong, Zhishang Liu. Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers). 2024.

The problem of traffic speed predictions is still very challenging due to the complex and dynamic urban traffic conditions. Many existing works have implied the importance of integrating spatial correlations into models to explore nonlinear spatio-temporal dependencies and make traffic predictions in near future. However, some of the works only consider pairwise correlations and cannot model the hidden information among multiple nodes well, and the others only consider hyper-correlations (that can be shared by more than two nodes) and discount the role of the pairwise ones for propagating spatial dependencies. Therefore, we propose a Spatio-Temporal neural nEtwork based on both Pairwise and Hyper-correlations (STEPH) for traffic speed predictions. It is distinguished primarily by incorporating both types of spatial correlations into temporal information and designing new hybrid spatio-temporal blocks in neural networks to effectively overcome the challenge. Experiments on two real-world traffic datasets demonstrate the effectiveness of the proposed model, and show its superiority performance compared to other state-of-the-art baselines.

Recommendations are important web services in the era of information explosion. Particularly, group recommendations aim to suggest new items to groups such that the members of groups are likely interested in. However, existing works still suffer from sparsity and cold-start issues (e.g., cold-start groups or items) for groups with few interactions on items. Most of them model the preferences or features of entities (i.e., users, items and groups) from heterogeneous interactions (i.e., user-item, group-item and user-group interactions) between two distinct types of entities, while ignoring the homogeneous interactions (i.e., user-user, item-item and group-group interactions) between entities of one type. To this end, we propose a new model, called H3Rec, which learns the representations of entities by developing two graph embedding layers based on an interaction graph of all entities. Specifically, the two graph embedding layers make full use of the hidden information in theHigher-orderHeterogeneous andHomogeneous interactionsof the graph. Therefore, H3Rec can alleviate the sparsity and cold-start issues and improve the performance of group recommendations. The experimental results on two real world datasets in different domains show the superiority of H3Rec in group recommendations, especially for cold-start groups and items.

Traffic is very important to route planning and people’s daily lives. Traffic prediction is still very challenging as it is affected by many complex factors including dynamic spatio-temporal dependencies and external factors (e.g., road types and nearby points of interest) in the road network. Dynamic spatio-temporal dependencies simultaneously contain spatial and temporal dependencies. Existing models for predicting traffic of links only consider the spatial dependencies from the perspective of links or the whole road network by ignoring the spatial dependencies among regions. To this end, this paper proposes a new Spatio-Temporal Neural Network (STNN) with the encoder-decoder architecture to improve the accuracy of traffic predictions by additionally taking into account the region-based spatial dependencies and external factors. Specifically, STNN learns dynamic spatio-temporal dependencies from historical traffic time series via an encoder in the perspective of the road network, with two spatial models, i.e.,region-based spatial modelandlink-based spatial attention modelin the perspectives of regions and links, respectively. Further, STNN decodes the output from the encoder via a decoder with a temporal attention model for recording long-term dependencies and fuses external factors in the road network, to improve network-wide traffic predictions. We conduct extensive experiments to evaluate the performance of STNN on three real-world traffic datasets, which shows that STNN is significantly better than the state-of-the-art models.

The sequential pattern in the human movement is one of the most important aspects for location recommendations in geosocial networks. Existing location recommenders have to access users' raw check-in data to mine their sequential patterns that raises serious location privacy breaches. In this paper, we propose a new Privacy-preserving LOcation REcommendation framework (PLORE) to address this privacy challenge. First, we employ the nnth-order additive Markov chain to exploit users' sequential patterns for location recommendations. Further, we contrive the probabilistic differential privacy mechanism to reach a good trade-off between high recommendation accuracy and strict location privacy protection. Finally, we conduct extensive experiments to evaluate the performance of PLORE using three large-scale real-world data sets. Extensive experimental results show that PLORE provides efficient and highly accurate location recommendations, and guarantees strict privacy protection for user check-in data in geosocial networks.

Nowadays, many researchers study on characterizing complex and dynamic traffic environments by modeling the spatio-temporal dependencies in a road network for traffic predictions. However, existing works fail to investigate comprehensive spatio-temporal dependencies, because most of them ignore the spatial dependencies from the topological graph structure information in the road network, or only consider the temporal dependencies between fine-grained time slots whereas ignore those among coarse-grained time periods, in which a time period is composed of a number of time slots. To this end, we propose a new encoder-decoder framework called GAMIT with graphical space and multi-grained time by developing a spatiotemporal recurrent neural network (STRNN) for traffic predictions, where the graphical space consists of spatial networks which represent road networks. STRNN first devises a spatiotemporal convolution block to capture the fine-grained spatiotemporal dependencies between time slots in the road network. Then, STRNN uses the recurrent architecture to catch the coarsegrained spatio-temporal dependencies among time periods. The encoder finally applies STRNN to learn the multi-grained spatiotemporal dependencies which are fed into the decoder for computing traffic predictions based on STRNN as well. To evaluate the performance of GAMIT, we conduct extensive experiments on two real traffic flow datasets. Experimental results show that GAMIT outperforms the state-of-the-art traffic prediction models.

Point-of-interest (POI) recommendation has become an important service to help users discover attractive locations. A variety of available check-in data make it possible to build a personalized POI recommender system, but the extreme sparsity of check-in data poses a severe challenge for POI recommendation. Recent studies mainly utilize social information, categorical information and/or geographical information to supplement the highly sparse check-in data. However, these studies often apply shallow methods for the extra information and provide considerably limited improvements on POI recommendation. In this paper, we propose a fine-grained POI recommendation framework, called FGRec to capture the intrinsic influences of social, categorical and geographical information on the check-in behaviors of users. First, we study the social influence in depth by exploiting the multi-hop social friends and top-n nearest neighbor friends, not only the direct friends (i.e., 1-hop friends). Second, we investigate the categorical influence by factorizing both user-POI and user-category matrices simultaneously over the same user embedding space, rather than simply using the popularity of POI categories. Third, we explore the geographical influence by integrating two types of distance (i.e., the distance between user homes and POIs and the distance among POIs) into a unified probability distribution over check-in POIs, instead of modeling them separately. Finally, experimental results on two large-scale real-world datasets demonstrate the effectiveness and superiority of the proposed method.

Group recommendation aims to suggest preferred items to a group of users rather than to an individual user. Most existing methods on group recommendation directly learn theinherent interests of groups and users orinherent features of items, i.e., independently modeling the inherent embeddings of groups, users or items. However, the independent view severely suffers from the cold-start problem when making recommendations for occasional groups that are temporally formed by a set of users and have few interactions on items. Actually, the groups, users and items are interdependent because they interact with one another. The interdependencies constitute an interaction graph that provides multiple views to model the embeddings of groups, users and items from their interacting counterparts to improve recommendation for occasional groups. To this end, we propose a model, named GAME to learn the Graphical and Attentive Multi-view Embeddings (i.e., representations) for the groups, users and items from the independent view and counterpart views based on the interaction graph. In the counterpart views, the embedding of a group, user or item is aggregated from the interacting counterparts based on an attention mechanism that derives the adaptive weight for each counterpart. For instance, a user's embedding may be aggregated from her interacting items or groups. Further, GAME applies neural collaborative filtering to investigate the interactions between the multi-view embeddings of groups (or users) and items for group recommendation. Finally, we conduct extensive experiments on two real datasets. The experimental results show that GAME outperforms other state-of-the-art models, especially on both cold-start groups (i.e., occasional groups) and cold-start items.

Most of current group recommenders only consider the bottom-up influences, i.e., the preference of a group is greatly affected by the members in the group. For example, children usually dominate the preference of a family while senior experts often lead the preference of a professional group. However, in reality there also exist the top-down influences, i.e., a group inherently affects its every member, because a group often has some distinct themes which limit the preferences of all the members in the group. For instance, the members in a group for sports may prefer hiking and rock climbing, whereas the members in a group for entertainment would like to watch movies and play games. In other words, the influences between a group and its members are dual. To this end, this paper proposes a new model for Group Recommendation using Attentive Dual Influences (GRADI) that simultaneously explores both the bottom-up and top-down influences between a group and its members. The preference of a member in a group is represented as the group-specific member embedding by modeling the top-down influences from the group to the member. In addition, the preference of a group on a target item is represented as the item-specific group representation by considering the bottom-up influences from all the members to the group, where an attentive mechanism is developed to aggregate the preferences of all the members on a target item. Furthermore, GRADI investigates the interactions between groups and items with neural collaborative filtering. Results of extensive experiments conducted on two real-world datasets show that GRADI outperforms other state-of-the-art models.

As many location-based applications provide services for users based on traffic conditions, an accurate traffic prediction model is very significant, particularly for long-term traffic predictions (e.g., one week in advance). As far, long-term traffic predictions are still very challenging due to the dynamic nature of traffic. In this paper, we propose a model, called Spatio-Temporal Convolutional Neural Network (STCNN) based on convolutional long short-term memory units to address this challenge. STCNN aims to learn the spatio-temporal correlations from historical traffic data for long-term traffic predictions. Specifically, STCNN captures the general spatio-temporal traffic dependencies and the periodic traffic pattern. Further, STCNN integrates both traffic dependencies and traffic patterns to predict the long-term traffic. Finally, we conduct extensive experiments to evaluate STCNN on two real-world traffic datasets. Experimental results show that STCNN is significantly better than other state-of-the-art models.

In recent years, there has been a tremendous increase in the popularity of event-based social networks which allow social and physical interactions among their members. One major challenge for their members is the difficulty of searching events that meet their preferences from a large number of upcoming events. To tackle this challenge, we propose a personalized event recommendation framework called EventRec that exploits the geographical, social and temporal influences of events on users to generate personalized event recommendations. In EventRec, we model the influence of two-dimensional geographical location of an event using the Kernel Density Estimation method along with the popularity of the event location. Furthermore, the social influence in EventRec does not rely only on the relevance of a group to a user, but it also considers the relevance of the group to her friends. The geographical and social influences are integrated with the temporal influence that considers the preferences of the user and her friends on the days of the week and time of events. Our performance evaluation is conducted using two large Meetup.com data sets, and experimental results show that the quality of recommendations of EventRec outperforms the state-of-the-art event recommendation techniques.

In recommender systems, one key task is to predict the personalized rating of a user to a new item and then return the new items having the top predicted ratings to the user. Recommender systems usually apply collaborative filtering techniques (e.g., matrix factorization) over a sparse user-item rating matrix to make rating prediction. However, the collaborative filtering techniques are severely affected by the data sparsity of the underlying user-item rating matrix and often confront the cold-start problems for new items and users. Since the attributes of items and social links between users become increasingly accessible in the Internet, this paper exploits the rich attributes of items and social links of users to alleviate the rating sparsity effect and tackle the cold-start problems. Specifically, we first propose a Kernel-based Attribute-aware Matrix Factorization model called KAMF to integrate the attribute information of items into matrix factorization. KAMF can discover the nonlinear interactions among attributes, users, and items, which mitigate the rating sparsity effect and deal with the cold-start problem for new items by nature. Further, we extend KAMF to address the cold-start problem for new users by utilizing the social links between users. Finally, we conduct a comprehensive performance evaluation for KAMF using two large-scale real-world data sets recently released in Yelp and MovieLens. Experimental results show that KAMF achieves significantly superior performance against other state-of-the-art rating prediction techniques.

Geosocial networks like Yelp and Foursquare have been rapidly growing and accumulating plenty of data such as social links between users, user check-ins to venues, venue geographical locations, venue categories, and user textual comments on venues. These data contain rich knowledge on the user's social interactions in communities, geographical mobility patterns between regions, categorical preferences on activities, aspect interests in topics, and opinion expressions for sentiments. Such knowledge is essential for two key applications, namely, text sentiment classification and venue recommendations, which will be developed in this paper. To extract the knowledge from the data, the key task is to discover the latent communities, regions, activities, topics, and sentiments of users. However, these latent variables are interdependent, e.g., users in the same community usually travel on nearby regions and share common activities and topics, which renders a big challenge for modeling these latent variables. To tackle this challenge, in this study, we propose an LDA-based model called CRATS that jointly mines the latent Communities, Regions, Activities, Topics, and Sentiments based on the important dependencies among these latent variables. To the best of our knowledge, this is the first study to jointly model these five latent variables. Finally, we conduct a comprehensive performance evaluation for CRATS in different applications, including text sentiment classification and venue recommendations, using three large-scale real-world geosocial network data sets collected from Yelp and Foursquare. Experimental results show that CRATS achieves significantly superior performance against other state-of-the-art techniques.

In location-based social networks (LBSNs), time significantly affects users' check-in behaviors, for example, people usually visit different places at different times of weekdays and weekends, e.g., restaurants at noon on weekdays and bars at midnight on weekends. Current studies use the temporal influence to recommend locations through dividing users' check-in locations into time slots based on their check-in time and learning their preferences to locations in each time slot separately. Unfortunately, these studies generally suffer from two major limitations: (1) the loss of time information because of dividing a day into time slots and (2) the lack of temporal influence correlations due to modeling users' preferences to locations for each time slot separately. In this paper, we propose a probabilistic framework called TICRec that utilizes temporal influence correlations (TIC) of both weekdays and weekends for time-aware location recommendations. TICRec not only recommends locations to users, but it also suggests when a user should visit a recommended location. In TICRec, we estimate a time probability density of a user visiting a new location without splitting the continuous time into discrete time slots to avoid the time information loss. To leverage the TIC, TICRec considers both user-based TIC (i.e., different users' check-in behaviors to the same location at different times) and location-based TIC (i.e., the same user's check-in behaviors to different locations at different times). Finally, we conduct a comprehensive performance evaluation for TICRec using two real data sets collected from Foursquare and Gowalla. Experimental results show that TICRec achieves significantly superior location recommendations compared to other state-of-the-art recommendation techniques with temporal influence.

The increasing prevalence of sensors and mobile devices has led to an explosive increase of the scale of spatio-temporal data in the form of trajectories. A trajectory aggregate query, as a fundamental functionality for measuring trajectory data, aims to retrieve the statistics of trajectories passing a user-specified spatio-temporal region. A large-scale spatio-temporal database with big disk-resident data takes very long time to produce exact answers to such queries. Hence, approximate query processing with a guaranteed error bound is a promising solution in many scenarios with stringent response-time requirements. In this paper, we study the problem of approximate query processing for trajectory aggregate queries. We show that it boils down to the distinct value estimation problem, which has been proven to be very hard with powerful negative results given that no index is built. By utilizing the well-established spatio-temporal index and introducing an inverted index to trajectory data, we are able to design random index sampling (RIS) algorithm to estimate the answers with a guaranteed error bound. To further improve system scalability, we extend RIS algorithm to concurrent random index sampling (CRIS) algorithm to process a number of trajectory aggregate queries arriving concurrently with overlapping spatio-temporal query regions. To demonstrate the efficacy and efficiency of our sampling and estimation methods, we applied them in a real large-scale user trajectory database collected from a cellular service provider in China. Our extensive evaluation results indicate that both RIS and CRIS outperform exhaustive search for single and concurrent trajectory aggregate queries by two orders of magnitude in terms of the query processing time, while preserving a relative error ratio lower than 10\%, with only 1% search cost of the exhaustive search method.

As location-based social networks (LBSNs) rapidly grow, it is a timely topic to study how to recommend users with interesting locations, known as points-of-interest (POIs). Most existing POI recommendation techniques only employ the check-in data of users in LBSNs to learn their preferences on POIs by assuming a user's check-in frequency to a POI explicitly reflects the level of her preference on the POI. However, in reality users usually visit POIs only once, so the users' check-ins may not be sufficient to derive their preferences using their check-in frequencies only. Actually, the preferences of users are exactly implied in their opinions in text-based tips commenting on POIs. In this paper, we propose an opinion-based POI recommendation framework called ORec to take full advantage of the user opinions on POIs expressed as tips. In ORec, there are two main challenges: (i) detecting the polarities of tips (positive, neutral or negative), and (ii) integrating them with check-in data including social links between users and geographical information of POIs. To address these two challenges, (1) we develop a supervised aspect-dependent approach to detect the polarity of a tip, and (2) we devise a method to fuse tip polarities with social links and geographical information into a unified POI recommendation framework. Finally, we conduct a comprehensive performance evaluation for ORec using two large-scale real data sets collected from Foursquare and Yelp. Experimental results show that ORec achieves significantly superior polarity detection and POI recommendation accuracy compared to other state-of-the-art polarity detection and POI recommendation techniques.

Recommending users with their preferred points-of-interest (POIs), e.g., museums and restaurants, has become an important feature for location-based social networks (LBSNs), which benefits people to explore new places and businesses to discover potential customers. However, because users only check in a few POIs in an LBSN, the user-POI check-in interaction is highly sparse, which renders a big challenge for POI recommendations. To tackle this challenge, in this study we propose a new POI recommendation approach called GeoSoCa through exploiting geographical correlations, social correlations and categorical correlations among users and POIs. The geographical, social and categorical correlations can be learned from the historical check-in data of users on POIs and utilized to predict the relevance score of a user to an unvisited POI so as to make recommendations for users. First, in GeoSoCa we propose a kernel estimation method with an adaptive bandwidth to determine a personalized check-in distribution of POIs for each user that naturally models the geographical correlations between POIs. Then, GeoSoCa aggregates the check-in frequency or rating of a user's friends on a POI and models the social check-in frequency or rating as a power-law distribution to employ the social correlations between users. Further, GeoSoCa applies the bias of a user on a POI category to weigh the popularity of a POI in the corresponding category and models the weighed popularity as a power-law distribution to leverage the categorical correlations between POIs. Finally, we conduct a comprehensive performance evaluation for GeoSoCa using two large-scale real-world check-in data sets collected from Foursquare and Yelp. Experimental results show that GeoSoCa achieves significantly superior recommendation quality compared to other state-of-the-art POI recommendation techniques.

K-anonymity has been used to protect location privacy for location monitoring services in wireless sensor networks (WSNs), where sensor nodes work together to report k-anonymized aggregate locations to a server. Each k-anonymized aggregate location is a cloaked area that contains at least k persons. However, we identify an attack model to show that overlapping aggregate locations still pose privacy risks because an adversary can infer some overlapping areas with less than k persons that violates the k-anonymity privacy requirement. In this paper, we propose a reciprocal protocol for location privacy (REAL) in WSNs. In REAL, sensor nodes are required to autonomously organize their sensing areas into a set of non-overlapping and highly accurate k-anonymized aggregate locations. To confront the three key challenges in REAL, namely, self-organization, reciprocity property and high accuracy, we design a state transition process, a locking mechanism and a time delay mechanism, respectively. We compare the performance of REAL with current protocols through simulated experiments. The results show that REAL protects location privacy, provides more accurate query answers, and reduces communication and computational costs.

Recommending to users personalized locations is an important feature of Location-Based Social Networks (LBSNs), which benefits users who wish to explore new places and businesses to discover potential customers. In LBSNs, social and geographical influences have been intensively used in location recommendations. However, human movement also exhibits spatiotemporal sequential patterns, but only a few current studies consider the spatiotemporal sequential influence of locations on users’ check-in behaviors. In this article, we propose a new gravity model for location recommendations, called LORE, to exploit the spatiotemporal sequential influence on location recommendations. First, LORE extracts sequential patterns from historical check-in location sequences of all users as a Location-Location Transition Graph (L 2 TG), and utilizes the L 2 TG to predict the probability of a user visiting a new location through the developed additive Markov chain that considers the effect of all visited locations in the check-in history of the user on the new location. Furthermore, LORE applies our contrived gravity model to weigh the effect of each visited location on the new location derived from the personalized attractive force (i.e., the weight) between the visited location and the new location. The gravity model effectively integrates the spatiotemporal, social, and popularity influences by estimating a power-law distribution based on (i) the spatial distance and temporal difference between two consecutive check-in locations of the same user, (ii) the check-in frequency of social friends, and (iii) the popularity of locations from all users. Finally, we conduct a comprehensive performance evaluation for LORE using three large-scale real-world datasets collected from Foursquare, Gowalla, and Brightkite. Experimental results show that LORE achieves significantly superior location recommendations compared to other state-of-the-art location recommendation techniques.

Geographical influence has been intensively exploited for location recommendations in location-based social networks (LBSNs) due to the fact that geographical proximity significantly affects users’ check-in behaviors. However, current studies only model the geographical influence on all users’ check-in behaviors as auniversalway. We argue that the geographical influence on users’ check-in behaviors should bepersonalized. In this paper, we propose a personalized and efficient geographical location recommendation framework called iGeoRec to take full advantage of the geographical influence on location recommendations. In iGeoRec, there are mainly two challenges: (1) personalizing the geographical influence to accurately predict the probability of a user visiting a new location, and (2) efficiently computing the probability of each user to all new locations. To address these two challenges, (1) we propose a probabilistic approach to personalize the geographical influence as a personal distribution for each user and predict the probability of a user visiting any new location using her personal distribution. Furthermore, (2) we develop an efficient approximation method to compute the probability of any user to all new locations; the proposed method reduces the computational complexity of the exact computation method from$O(|L|n^3)$to$O(|L|n)$(where$|L|$is the total number of locations in an LBSN and$n$is the number of check-in locations of a user). Finally, we conduct extensive experiments to evaluate the recommendationaccuracyandefficiencyof iGeoRec using two large-scale real data sets collected from the two of the most popular LBSNs: Foursquare and Gowalla. Experimental results show that iGeoRec provides significantly superior performance compared to other state-of-the-art geographical recommendation techniques.

Providing location recommendations becomes an important feature for location-based social networks (LBSNs), since it helps users explore new places and makes LBSNs more prevalent to users. In LBSNs, geographical influence and social influence have been intensively used in location recommendations based on the facts that geographical proximity of locations significantly affects users' check-in behaviors and social friends often have common interests. Although human movement exhibits sequential patterns, most current studies on location recommendations do not consider any sequential influence of locations on users' check-in behaviors. In this paper, we propose a new approach called LORE to exploit sequential influence on location recommendations. First, LORE incrementally mines sequential patterns from location sequences and represents the sequential patterns as a dynamic Location-Location Transition Graph (L2TG). LORE then predicts the probability of a user visiting a location by Additive Markov Chain (AMC) with L2TG. Finally, LORE fuses sequential influence with geographical influence and social influence into a unified recommendation framework; in particular the geographical influence is modeled as two-dimensional check-in probability distributions rather than one-dimensional distance probability distributions in existing works. We conduct a comprehensive performance evaluation for LORE using two large-scale real data sets collected from Foursquare and Gowalla. Experimental results show that LORE achieves significantly superior location recommendations compared to other state-of-the-art recommendation techniques.

Location-tagged social media have an increasingly important role in shaping behavior of individuals. With the help of location recommendations, users are able to learn about events, products or places of interest that are relevant to their preferences. User locations and movement patterns are available from geosocial networks such as Foursquare, mass transit logs or traffic monitoring systems. However, disclosing movement data raises serious privacy concerns, as the history of visited locations can reveal sensitive details about an individual's health status, alternative lifestyle, etc. In this paper, we investigate mechanisms to sanitize location data used in recommendations with the help of differential privacy. We also identify the main factors that must be taken into account to improve accuracy. Extensive experimental results on real-world datasets show that a careful choice of differential privacy technique leads to satisfactory location recommendation results.

A temporary social network (TSN) is confined to a specific place (e.g., hotel and shopping mall) or activity (e.g., concert and exhibition) in which the TSN service provider allows nearby third party vendors (e.g., restaurants and stores) to advertise their goods or services to its registered users. However, simply broadcasting all the vendors' advertisements to all the users in the TSN may cause the service provider to lose its fans. In this paper, we present Capacity-Aware Location-Based Advertising (CALBA), which is a framework designed for TSNs to select vendors as advertising sources for mobile users. In CALBA we measure the relevance of a vendor to a user by considering their geographical proximity and the user's preferences. Our goal is to maximize the overall relevance of selected vendors for a user with the constraint that the total advertising frequency of the selected vendors should not exceed the user's specified capacity. First, we model the snapshot selection problem as 0-1 knapsack and solve it using an approximation method. Then, CALBA keeps track of the selection result for moving users by employing a safe region technique that can reduce its computational cost. We also propose three pruning rules and a unique access order to effectively prune vendors which could not affect a safe region, in order to improve the efficiency of the client-side computation. We evaluate the performance of CALBA based on a real location-based social network data set crawled from Foursquare. Experimental results show that CALBA outperforms a naïve approach which periodically invokes the snapshot vendor selection.

With the rapidly growing location-based social networks (LBSNs), personalized geo-social recommendation becomes an important feature for LBSNs. Personalized geo-social recommendation not only helps users explore new places but also makes LBSNs more prevalent to users. In LBSNs, aside from user preference and social influence, geographical influence has also been intensively exploited in the process of location recommendation based on the fact that geographical proximity significantly affects users' check-in behaviors. Although geographical influence on users should be personalized, current studies only model the geographical influence on all users' check-in behaviors in a universal way. In this paper, we propose a new framework called iGSLR to exploit personalized social and geographical influence on location recommendation. iGSLR uses a kernel density estimation approach to personalize the geographical influence on users' check-in behaviors as individual distributions rather than a universal distribution for all users. Furthermore, user preference, social influence, and personalized geographical influence are integrated into a unified geo-social recommendation framework. We conduct a comprehensive performance evaluation for iGSLR using two large-scale real data sets collected from Foursquare and Gowalla which are two of the most popular LBSNs. Experimental results show that iGSLR provides significantly superior location recommendation compared to other state-of-the-art geo-social recommendation techniques.