Successive Point-of-interest Recommendation In Location-based Social Networks

Recent years have witnessed the rapid development of online social networks,in which Internet users access to vast amount of information on an unprecedented scale.Advances in wireless communication technologies and location techniques enhance the social networking services,allowing users to check-in venues online and share their experiences towards pointof-interest(POIs)in the physical world via their mobile devices.These social networks are referred to as location-based social networks(LBSNs),such as Foursquare,Gowalla,Facebook Place and Geo Life,and there has been an increasing enthusiasm on developing the LBSNs.LBSNs bridges the gap between the physical and digital worlds,and check-in information can be used to study life patterns of LBSN users and mine their preferences on locations.In order to improve user experiences in LBSNs,POI recommendation is proposed to suggest interesting places to users by mining the check-in records and contextual information in LBSNs.POI recommendation has become a popular research issue and attracted much effort from both academia and industry.However,traditional studies on POI recommendation overlooked the consecutive behaviors of each user,which is an important feature to leverage on as human mobility often exhibits sequential patterns.We argue that an ideal POI recommender system should provide prompt suggestions with respect to users' current location.Some recent works started to address the aforementioned issue as next POI recommendation(also called successive POI recommendation)to predict a user's very next move.Next POI recommendation not only helps users to explore their favorite places but also benefits to advertising agency to provide an effective way of launching advertisement to target the potential clients.However,to achieve accurate next POI recommendation is very challenging,since the check-in is a type of implicit feedback and the check-in data of each user is highly sparse.To this end,this thesis aims to research on next POI recommendation models and algorithms by employing various context information,and provide more effective location-based services to users.The main contributions and innovations are summarized as follows:(1)Inferring a personalized next POI recommendation model with latent behavior patterns.With the conjecture that,under different contextual scenario,the human exhibit distinct mobility patterns,we attempt here to jointly model the next POI recommendation under the influence of user's latent behavior pattern.In this work,we propose to adopt a third-rank tensor to model the successive check-in behaviors.By incorporating softmax function to fuse the personalized Markov chain with latent pattern,we furnish a Bayesian Personalized Ranking(BPR)approach and derive the optimization criterion accordingly.In the model learning phase,the Expectation Maximization(EM)is used to estimate the model parameters.Extensive experiments on three large-scale real-world LBSNs datasets demonstrate the significant improvements of our model over several state-of-the-art methods.(2)Category-aware next POI recommendation.The categorical information is useful for modeling users' preference on POIs,and many recent works had tried to take categorical information into account to improve the recommendation performance.However,previous efforts suffer from the high computational complexity,besides the transition pattern between POIs has not been well studied.In this work,we proposed a two-fold approach for next POI recommendation.First,we predict the category of next location by using a third-rank tensor optimized by a Listwise Bayesian Personalized Ranking(LBPR)approach.Specifically,we introduce two functions,namely Plackett-Luce model and cross entropy,to generate the likelihood of a ranking list for posterior computation.Then POI candidates filtered by the predicated category are ranked based on the spatial influence and category ranking influence.By explicit usage of category information to infer the user transition pattern in category-level,the proposed models are perfect for the next new POI recommendation problem.The experiments on two real-world datasets demonstrate the significant improvements of our methods over several state-of-the-art methods.(3)A personalized next point-of-interest recommendation with transition interval patterns.Temporal information plays an important role in POI recommendations and most existing studies apply the observed check-in time stamps explicitly to improve the recommendation performance.With the conjecture that the transition intervals between two successive check-ins may carry more information for diversified behavior patterns,we propose a probabilistic factor analysis model to incorporate three components,namely,personal preference,distance preference,and transition interval preference.They are modeled by an observed third-rank transition tensor,a distance constraint,and a continuous latent variable,respectively.In this model,the POI recommendation and the transition interval for user's very next move can be inferred simultaneously by maximizing the posterior probability of the overall transitions.Expectation Maximization(EM)algorithm is used to tune the model parameters.Experimental results demonstrate that the proposed model outperforms other state-of-the-art methods in terms of next POI recommendation and the expected transition time.