Research On Location Recommendation Algorithm Based On LBSN

With the development of social networking and location technologies,locationbased social networks are growing in size.People can connect with friends,explore interesting places and share their places and experiences in a check-in manner on a location-based social networks.With the massive growth of check-in data,the value of data is growing.How to filter out valuable information from this huge data and provide users with personalized recommendations has become the focus of scholars' research.Among them,according to the check-in data of the user history,recommending the location of interest to the user is a popular research direction.This paper studies the related algorithms recommended by the location and finds that many previous LBSN-based collaborative filtering algorithms do not consider the user's current location.Therefore,no matter where the user is,the recommendation algorithm will recommend the same location to the user,which will result in the recommended location being far from the user's current location.Consider a scenario where the user leaves his living area and travels to a faraway place.Based on the recommendation algorithm that did not consider the user's current location before,the location recommended to the user will be close to his residential area.Because the user's historical check-in data is near the residential area,the recommended location based on the user's collaborative filtering algorithm will also be close to the user's living area,and the location near the user's travel location will not be recommended.Due to the limitations of time and space,the recommended location away from the user's current location is of no practical significance.In this paper,an in-depth study is carried out on this problem.From the three aspects of user similarity,social factors and geographical factors,the user-based collaborative filtering algorithm is improved,and a distance adaptive algorithm is proposed.There are two simple but important features of this algorithm.1)The algorithm filters out far-reaching candidate locations based on the distance between the recommended location and the user's current location.2)The algorithm integrates the time factor into the calculation of user similarity,and integrates the familiarity of friends into social factors.This paper uses the dataset in Foursquare to compare and analyze the improved algorithm with other recommended algorithms,and verifies the feasibility of the improved algorithm.The experimental results show that the improved algorithm improves the recommendation effect to some extent.