Reseaech And Implementation Of Region Aware POI Recommendation Technology Driven By Knowledge Graph

With the rapid development of mobile Internet,location information service applications represented by Dianping and Yelp have become increasingly popular.Users share their favorite POIs through check-ins,comments and other behaviors to record what they see,hear and travel footprints.Due to the exponential growth of POI data,in order to help users find the locations of interest,POI recommendation technology came into being.Based on the results of POI recommendation,service providers can help users quickly obtain POIs that satisfy their preferences and improve their experience;Meanwhile,they can also help stores accurately attract potential customers and increase marketing revenue.Compared with recommendation tasks such as movies and music,POI recommendation tasks face severe challenges due to their strong data sparseness and large geographic-related influence.Related research shows that incorporating semantic influences(such as attributes,categories,etc.)and geographic influences into the recommendation model can improve the performance of POI recommendation.However,current related research still has certain shortcomings in considering the semantic relevance of POIs and the granularity of geographical regions.On the one hand,users usually like POIs with similar categories and attributes,but existing research work usually uses categories as filtering conditions.Accurate matching of categories is required,and the semantic relevance of between similar categories of POIs is not fully explored.On the other hand,existing work usually divides the city into fixed-size regions when considering geographic influences,but coarse-grained division will affect the accuracy of recommendation,while fine-grained division will lead to high computational overhead,sparse samples,which cannot adapt to real-time POI recommendation scenarios with dynamic changes in regional environments.For this reason,this thesis introduces using the knowledge graph into POI recommendation to explore the relevance of POIs,constructs the spatial semantic graph and characterizes the dynamic region environment,to achieve real-time POI recommendation.The specific research work of this thesis mainly includes:(1)Propose a POI recommendation model based on the semantic relevance of knowledge graph.This thesis studies the use of knowledge graphs to mine the relevance of POIs,and proposes a semantic feature extraction mechanism for POIs based on knowledge graphs.In view of the different characteristics of the attributes and categories of POIs,the graph representation learning technology and Poincaré spherical model are used to obtain semantic representations of the implicit POI relevance,so as to design the POI recommendation model.The effectiveness of the proposed model is verified on real datasets.(2)Propose a real-time POI recommendation model for dynamic regions.This thesis constructs a spatial semantic graph based on the POI knowledge graph,proposes a dynamic regional feature characterization mechanism based on the spatial semantic graph and learns the semantic feature aggregation method of POIs to adaptively characterize regional environmental features in different geographic regions;And based on the dynamic region feature modeling in the space,the user's long-short term interest preferences that change dynamically in time series are considered,so as to realize the real-time POI recommendation for dynamic regions.The effectiveness of the proposed model is verified on real datasets.(3)Design and implement a real-time POI recommendation prototype system based on knowledge graphs.The offline part of the system is responsible for data management and model training,and the online part of the system is responsible for personalized display of POI recommendation results that meet real-time requirements.Verify the real-time performance of the recommendation effect on the system performance.In summary,this thesis introduces using the knowledge graph into POI recommendation.By constructing the POI knowledge graph and the spatial semantic graph,real-time POI recommendation oriented to semantic relevance and dynamic regional environment is realized.Based on the theoretical research results,a prototype system for real-time POI recommendation based on knowledge graph is designed and implemented.The results of this research will promote the application of knowledge graphs in urban computing,and provide support for the construction of POI recommendation mechanisms and related local life service applications.