Research On POI Recommendation Algorithms Combining User Preferences And Geographical Influence

The popularity of mobile network and the maturity of GPS technology make the location-based social networks(LBSNs)develop rapidly,compared with traditional social networks,users in LBSNs can share their dynamics with tagging their location in the way of check-in.The accumulation of users' check-in records provides data support for analyzing users' behavior habits and mining users' location preferences.Point-of-interest(POI)recommendation has become one of the important tasks in location-based social networks.POI recommendation can not only recommend POIs to users and enrich users' lives,but also promote businesses and attract more consumers.However,due to the complex environment of location recommendation,user behavior is influenced by many factors,user-location check-in matrix is sparse and insufficient user preference mining,it is of great theoretical and practical significance to study POI recommendation.The main work of this paper is as follows.(1)In this paper,the status quo of location recommendation research at home and abroad is analyzed in detail,and the classification and related theory technology of location recommendation are analyzed.In view of the shortcomings of existing location recommendation algorithms,the research ideas and contents of this paper are put forward.(2)Based on the user-based collaborative filtering algorithm,a standardized check-in frequencies is proposed to represent the user POI preference.Then,according to the regional characteristics of the users' check-in activity,a fusion geographical influence model is proposed,which combines geographical distance and regional features to model geographical influence.Finally,a fusion model UGR is proposed and the experimental results on Gowalla dataset show that UGR algorithm can effectively improve the accuracy and recall of recommendation results.(3)POI category is an important location information,which can clearly reflect the functional and semantic information of the POI.Similar to the correlation between successive check-in POIs,there is also a correlation between categories of successive check-in POIs,and it can avoid the data sparsity problem by directly using POIs.Firstly,the user's location category correlation matrix is constructed,and the correlation of location category pairs is calculated by matrix decomposition algorithm.Then,a category-aware model CAR is proposed based on user preferences and geographical influence.The experimental results on Foursquare dataset show that the CAR effectively improves the accuracy and recall rate of recommendation results.(4)On the basis of the research in this paper,a prototype system of POI recommendation based on user preferences and geographic influence is designed and implemented.The overall design of the system and the functional modules are described.