Research On Multi-factor Point Of Interest Recommendation Based On Collaboratio

In recent years,with the development of the Internet,people are interacting more closely with smart mobile devices,making life more convenient,but also ushering in challenges.How can users quickly and accurately find the content they are really interested in from the complex data.The recommendation system can filter information for users and is widely used in many service platforms.In particular,location-based social network(LBSN)has grown tremendously with the increasing demand for personalized services.Point-of-interest(POI)recommendation has also developed rapidly as an important service and become one of the most popular research topics in LBSN.Most of the traditional POI recommendation methods are based on collaborative filtering techniques.However,since the user’s check-in data is an implicit feedback data in POI recommendation,these methods usually suffer from data sparsity problem.This problem can be alleviated by including other relevant supporting information in the modeling.For example,the user’s social relationship,the sequence information of interest points in the user’s check-in record,the distance between interest points,and the checkin time.Analysis and mining of this information allows for more accurate modeling of user preferences.There has been a lot of work on POI recommendation research using contextual information,which has driven the development of POI recommendation.POI recommendation studies have yielded promising results in location-based social networks.However,there are some problems with these methods: 1)they simply consider the influence of a single factor on users’ preferences and do not fully explore this factor;2)they consider multiple factors in modeling but ignore the constraints and coordination between multiple factors;3)they ignore the balance between long-term and short-term preferences of users.In this paper,we conduct an in-depth study on collaborative-based multi-factor point-of-interest recommendation methods and propose three effective POI recommendation models.First,this paper proposes a social relationship collaboration for point-of-interest recommendation model.The model models user preferences based on matrix decomposition,defining users with whom the user has explicit social relationships as display social friends and users with whom the user has common check-in POIs as implicit social friends.Calculate the closeness between users and their explicit and implicit friends,respectively.It also uses an adjustable parameter to balance the influence of explicit and implicit socialization on user preferences,from which the valid information in social relationships is fully explored.Experimental validation of the proposed method demonstrates the importance of dual social relationships for POI recommendations due to other contrasting recommendation methods.Secondly,this paper considers the integration of multiple factors into interest point recommendation and proposes a multi-factor collaborative interest point recommendation model based on user perception.We use matrix decomposition for modeling based on user perception,and use contextual information such as sequence,geography and social to construct a POI recommendation model with synergistic influence of user perception to alleviate the data sparsity problem.At the same time,in order to balance the mutual constraints between various factors,a multi-factor fusion strategy for user perception is proposed,which makes rational use of various contextual information to tap into the dynamic preferences of users.The experimental results prove that the proposed model in this paper has better performance in interest point recommendation.Finally,this paper proposes a graph-based multi-factor collaborative interest point recommendation model.The user check-in sequence and social relationship are constructed as a graph,and the aggregation and propagation of node information in the graph are performed using gated graph neural networks.Mining the spatial structure information in sequence information and the connection relationship between users in social relationship,respectively.Combining social relationship-based user representation and sequence information-based user representation.Then the hierarchical attention network is used to obtain users’ long-and short-term preferences,and further adaptively fuse the two to obtain users’ final preferences for model prediction.The experiments demonstrate that the model proposed in this paper obtains an improvement in the evaluation metrics in each dataset.