| In recent years,how to use the users’ historical check-in information to recommend point-of-interest that users are likely to visit has become one of the hot topics in the recommendation system.Users use social software based on location-based social networks to publish their likes and dislikes for a certain place,so historical check-in data can reflect users’ interest preferences.The historical check-in data of users generated by location-based social networks are sparse on the whole,and lacks the historical check-in data of new users.This makes it difficult to mine users’ check-in preferences in history,which in turn makes it difficult to recommend points of interest.Therefore,how to improve the accuracy of recommendation and fully mine the characteristics of users’ check-in behavior in the historical check-in data are the current problems to be solved.In order to solve the above problems,this paper uses knowledge graph and recurrent neural network and other related theories to study.First of all,in order to solve the problem of sparsity of data and insufficient logical and hierarchical information mining in the activities of users visit locations they want to go,this article put forward a point-of-interest recommendation algorithm based on knowledge graph.The algorithm makes full use of auxiliary data of check-in dataset by establishing knowledge graph based on LBSNs to reduce the influence of data sparsity.At the same time,the algorithm establishes a layer transfer mechanism integrating attention mechanism and hierarchical structure information,enriches the potential semantic information of users and interest points,and captures the logical and hierarchical information of users’ check-in activities.Then,a top-K recommendation list of points of interest that the user may visit is obtained based on the final user preference.Secondly,in order to solve the problem that the pre-existing point-of-interest recommendation algorithms ignore the spatial-temporal influence factors,and the temporal and spatial characteristics of user check-in behavior are not sufficient,this article put forward a point-of-interest recommendation algorithm based on recurrent neural network.Based on the constructed LBSNs knowledge graph,the algorithm mines the potential semantics of users and interest points.At the same time,the algorithm establishes the user check-in sequence according to the user check-in record and gets the temporal and spatial aspect of users check-in behavior.The algorithm takes advantage of the encoder-decoder structure based on the recurrent neural network to capture the long and short term dependence of user check-in behavior,and fully excavates the temporal and spatial preference of user check-in behavior.Then the user’s temporal and spatial preference is finally obtained,and the top-K recommendation list of interest points that the user may visit is obtained.Finally,experiments are carried out on Foursquare and Gowalla datasets to verify the performance of the proposed two point-of-interest recommendation algorithms.Through comparative analysis,the effectiveness of the proposed algorithms in improving recommendation quality is verified. |
