Point-of-Interest Recommendation Alogrithms Combing Context In Location-based Scoial Network

With the rapid development of Web 2.0,location acquisition and wireless communication technologies,a number of location-based social network(LBSN)have emerged and flourished,such as Foursquare,and Facebook Places,where users can check in at Point-of-Interests(POIs)and share life experiences in the physical world via mobile devices.On one hand,the new dimension of location implies extensive knowledge about an individual's behaviors and interests by bridging the gap between online social networks and the physical world,which enables us to better understand user behaviors and user preferences.On the other hand,it is valuable to develop the POI recommendation service as an essential function of LBSN to encourage users to explore new locations and help advertising agencies to launch location recommendation for potential customers and improve business profits.As a result,in recent years,the POI recommendation systems have received an increasing amount of attention from the academic and industrial fields.POI recommendation is a branch of recommendation systems,in which the POIs are treat as the items in the traditional recommendation system.We suffice to make use of conventional recommendation system techniques,e.g.,collaborative filtering methods.However,the specific fact that the location concatenating the physical world and the online networking services has its unique properties,structures and characteristics,which arouses new challenges to the traditional recommendation system techniques.We summarize some confronting challenges as follows.1.Complex relations:A location is a new object in location-based social network(LBSN),generating new relations between users,between locations,between users and times,and between users and locations,which makes the correlations between the above objects more complex.New recommendation scenarios,like POI recommendations,can be enabled using this new knowledge to enhance the performance.Therefore,it is necssarry to find out how to fuse these complex relations to generate high-quality recommendations.2.Rich context information:A location is one of the most important components defining a user's context.Moreover,LBSN also consist of different kinds of information,including not only check-in records,the geographical information of locations,and venue descriptions but also users' social relation information and media information(e.g.,user comments and videos).The heterogeneous information depicts the user activity from a variety of perspectives,helping the POI recommendation system to accurately mine user preferences and better understand user check-in behavior.Therefore,it is necessary to investigate how to fuse these heterogeneous data to produce high quality recommendations.To tackle these challenges,this paper proposes a series of new POI recommendation algorithms that integrate rich contextual information and complex relations to improve the recommendation performance.Specifically,the main works and contributions of this paper are as follows:1.Current researches on POI recommendation are generally limited to:(1)when modeling geographical influence,users' personalized behavior differences are ignored;(2)when modeling the users' social influence,the implicit social influence is seldom exploited.In this paper,we propose a novel POI recommendation approach called GeoEISo,which achieves three key goals in this work:(1)we develop a kernel estimation method with a self-adaptive kernel bandwidth to model the geographical influence among POIs.(2)we use the Gaussian radial basis kernel function based support vector regression(SVR)model to predict explicit trust values between users,and then devise a novel trust-based recommendation algorithm to simultaneously incorporate both the explicit and implicit social trust information into the process of POI recommendation.(3)We devise a unified geo-social framework which integrates users' preference on a POI with the geographical influence as well as social correlations.Experimental results on two real-world datasets collected from Foursquare show that GeoEISo provides significantly superior performances compared with other state-of-the-art POI recommendation algorithms.2.In order to tackle the problem of data sparity in POI recommendation,most existing models of POI recommendation on location-based social network(LBSN)improve recommendation quality by exploiting contextual information(e.g.,geographical information and social correlations).However,they tend to ignore the review texts information accompanied with rating information for recommender algorithms.To overcome this shortcoming,a novel POI recommendation algorithm called GeoSoRev is proposed in this paper.Specifically,we adopt a topic modeling technique(i.e.Non-negative Matrix Factorization(NMF))to model the latent topics in review texts.Then we integrate geographical information,social correlations and reviews text based on matrix factorization for POI recommedation.Experimental results on two real-world datasets collected from Foursquare show that GeoSoRev achieves significantly superior when compared with other state-of-the-art POI recommendation algorithms in terms of precision and recall.3.Nowadays,most researches focus on fitting a preference scoring function based on users' checking-in behaviors.The fact is,however,relying solely on the rating fails to reflect the user's preferences very accurately,because the users are most concerned with the list of ranked POIs on the actual output of recommender systems.In this paper,we propose a pairwise ranking algorithm called Geo-Social Bayesian Personalized Ranking algorithm(GSBPR),which is based on the pairwise ranking with the exploiting geo-social correlations by introduces technologies in the field of learning to rank into the process of POI recommendation.In this algorithm,we propose a novel BPR pairwise ranking assumption based on geo-social feedback by injecting users' geo-social preference.Based on this assumption,the POI recommendation problem is reformulated by a three-level joint pairwise ranking scheme.And the experimental results based on real datasets show that the proposed method in this paper enjoys better recommendation performance compared to other state-of-the-art POI recommendation algorithms.4.In reality,when providing the POI recommendations for users,spatial-temporal sequence of locations on users' check-in behaviors have been intensively studied in location recommendation based on the fact that the human movement exhibits a pattern of spatial-temporal sequence.However,current researches mainly focus on the spatial correlation among the sequences of users' checked-in locations,resulting in the following problems:(1)all previous works ignore the fact that location recommendation is a time-subtle recommendation task.In fact,users would prefer different check-in sequences on different time;(2)the influence of social information between users on the final recommendation results is ignored;(3)most works ignore that users' check in feedback is implicit and only positive examples are observed.To solve the problems mentioned above,a new location recommendation algorithm called Spatial-Temporal aware Social Collaborative Ranking(STSCR)algorithm is proposed to explore the impact of time,spatial-temporal sequential influence and social influence.In particular,the proposed algorithm is built upon a unified tensor factorization framework in which the fine-grained modeling of user-location,user-friend,friend-location,location-time,and location-location interactions.Then,the Bayesian personalized ranking technique is used to optimized the loss function of unified tensor factorization and fit the partial order of locations.As is shown in the experimental results on real datasets,our proposed STSCR achieves better recommendation performances than the other state-of-the-art location recommendation algorithms.