Research On Points Of Interest Recommendation In Social Media For Multi-Scenarios

With the rapid development of location-based social networks(LBSNs),point of interests(POI)recommendation has become an important means to help people discover attractive and interesting locations from billions of locations globally.However,this recommendation is very challenging compared to the traditional recommender systems.A user may visit only a limited number of POIs,leading to a very sparse user-item matrix.This matrix becomes even sparser when the user travels to a distant place as most of the items visited by a user are usually located within a short distance from the user's home.Moreover,user interests and behavior patterns may vary dramatically across different time and different geographical regions.On the other hand,in reality,human movement exhibits sequential patterns.Thus,how to predict users' next move based on her previous visited locations is important and challenging in LBSNs.Our project focuses on offering a more accurate and efficient recommender system by overcoming the aforementioned challenges,and it contains the following three parts:Personalized POI recommendation is crucial for LBSN,which not only helps users explore places but also enables many location-based services,e.g.,the targeting of mobile advertisements to users.However,personalized POI recommendation is highly challenging.LBSNs involve heterogeneous types of data,and the user-POI matrix is very sparse.To address these challenges,we analyze users' check-in behaviors in detail and introduce the concept of a heterogeneous information network(HIN).Then,we employ a meta path based approach to model geographical and social influences and a weighted meta path approach to model temporal influences.We propose an HIN-based POI recommendation system,which consists of two components: an improved singular value decomposition(SVD++)and factorization machines(FMs).For the similarity matrices that are generated by each meta path,we perform SVD++ to generate latent features for both users and POIs.With various meta-path-based features,we exploit FM with the group lasso to automatically select features.The results of experiments on two real-world LBSNs,namely,Gowalla and Foursquare,demonstrate that our method outperforms baseline methods in terms of accuracy.Next POI recommendation refers to predicting the next point of interest that the user will visit within a certain period of time.User travel behavior is affected by heterogeneous contextual factors,including continuous values(for example,geographic distance,time interval)and discrete values(for example,social status,weekly status).This chapter models multiple types of user behaviors based on multi-task learning.At the same time,because the difference between tasks in multi-task learning will damage the effect of some tasks,this chapter adopts a multi-door hybrid expert model based on sparse sharing structure to alleviate this problem.Experimental results based on real data sets show that compared with benchmark methods,the next point of interest recommendation algorithm proposed in this chapter has a significant improvement in effect.As more and more people organize gatherings online,recommendation of event points of interest for user groups becomes more and more important.Group-oriented(user group)event point of interest recommendation involves complex interactions between multiple entities(such as users,groups,events,points of interest,etc.).This chapter proposes an event interest point recommendation algorithm based on heterogeneous information network and deep neural network,which cleverly considers various factors.First,priority-based sampling techniques are used to select high-quality path instances.Then,the embedded representations of groups,events,points of interest,and meta-pathbased context are constructed,and a common attention mechanism is used to improve them.Based on the experimental results of real data sets,the effectiveness of the method in this chapter is verified.At the same time,the algorithm proposed in this chapter can recommend suitable points of interest for most popular events,and is very effective for recommending new points of interest.