Research On Point-of-interest Recommendation Algorithm Based On Multiple Data Sources In Location Social Networks

The development of network technologies and positioning technologies and the popularity of smart mobile devices have promoted the emergence and development of Location-based Social Networks(LBSN).In LBSN,users can make friends with other users and share the location information of their visits.Such locations are often referred to as point-of-interest(POIs).The POIs recommender is one of the most important applications in the location social network.Because the POIs recommender has great significance to users,businesses,and LBSN,it has aroused widespread concern in the academic community.Compared with the traditional recommender system,the POIs recommender has problems such as data sparse,user cold start problem,no explicit data.And it is influenced by many factors such as time and location.Multi-source data in the LBSN,such as time information,geographic location information,social information,content information etc.,provide a new ideas for POIs recommender.This article starts with the combination of multi-source data in LBSN,makes full use of multi-source data in LBSN,and improves the effectiveness of point of interest recommendations.This paper first summarizes the current domestic and foreign research status of current POIs recommender,briefly introduces the basic algorithms and commonly used algorithms in the current recommendation system field,analyzes the differences between the POIs recommender and the traditional recommendation items,and summarizes the difficulties in the POIs recommender.The analysis concludes that the multi-source data in the LBSN can be used to improve the effectiveness of the recommendation of interest points.This paper researches the POIs recommender based on multi-source data in LBSN.Based on this,we propose two recommender algorithms.The main tasks are as follows:1.We propose a Multi Probabilistic Matrix Factorization method of POIs recommender.This method first decomposes the three matrices using Multi Probabilistic Matrix Factorization,and combines the three aspects of information to obtain the user's potential features.Then,Gradient Boosting Decision Tree is used to train the features and tags to obtain the user's preference for interest points.And finally use the top-N recommender strategy considering the constraint problem generates a recommender list.Experimental results on real data sets show that the proposed method can achieve better results in terms of accuracy and F1 value than the currently popular methods,and can better alleviate the cold start problem in location services.Promote the recommendation of interest points.2.We propose a POIs recommender algorithm based on time characteristics and check-in sequence modeling.The method integrates time information with a check-in sequence,uses a word vector to model a check-in sequence,learns a vector representation of a point of interest,further calculates a user's vector representation,and then calculates a user's feature vector using multiple similarity calculations.The degree of similarity of the feature vectors of interest points is used to measure the user's preference for points of interest.Finally,experiments were conducted on real data sets,and compared with the benchmark algorithm,the effectiveness of the algorithm was proved.