Collective POI Recommendation Based On User Preference

Location-Based Service(hereinafter referred to as LBS)has been favored by the majority of users with the rapid development of mobile Internet in the past few years.Through the application of LBS,users can check in,comment,and share content at the Point of Interest(hereinafter referred to as POI).However,it has become one of the main problems that POI service providers and users pay close attention to,such as how to get the POI information which accords with the user's preference quickly and accurately from the massive data with the rapid development of LBS application.In this paper,we study the collective POI recommendation based on user preference,and the main work is as follows:1.In this paper,we propose a collective POI querying based on multiple keywords and user preference to solve the problem that the existing query methods do not consider the user preference and the best time to visit the POI.First of all,we present a problem named collective keywords preference query(hereinafter referred to CKPQ),which aims at finding a set of POIs covering all query keywords,satisfying user preference and having high accessibility.Then,we design a cost function to calculate the visit cost of candidate solution for the CKPQ problem.Finally,we propose an efficient query algorithm based on IR tree.In addition,the query space is reduced continuously through two pruning strategies,so as to improve the efficiency of the query algorithm.2.We propose a recommendation method of collective POIs based on the POI similarity and the personalized transition probability matrix to solve the problem of diversity of user's interest.As a matter of fact,a user's preference for a POI can be affected by a number of factors,such as the user's location and contextual information(such as time and geographic location,etc.).In this paper,we design an initial POI recommendation method based on rating prediction model,which considers the influence of user location,context information and POI popularity on the user's preference.Moreover,we propose the next POI recommendation method based on personalized transition probability matrix according to the initial POI.In this method,we construct the personalized transition probability matrix for each user with the consideration of trajectory similarity and time context information.Finally,we obtain the next POI by rating prediction model,and it is recommended to the user along with the initial POI.3.In this paper,we use real data to verify the two proposed methods.Aiming at the collective POI querying based on multiple keywords and user preference,we make a lot of experiments on the check-in data of 26520 POIs with 22389 users from Toronto extracted from Yelp.Thus,it proves that the proposed method is effective in running efficiency and query results.Aiming at the group POIs recommendation method based on POI similarity and personalized transition probability matrix,we extract the checkin information of 1463 POIs from 938 users in San Francisco from Foursquare and carries out experiments.As a result,it proves that the proposed method has higher accuracy than the contrast method.