Author : Rifeng Ding
Release : 2015
Genre :
Kind : eBook
Book Rating : /5 ( reviews)
Book Synopsis A Community-based Location Recommendation System for Location-based Social Networks by : Rifeng Ding
Download or read book A Community-based Location Recommendation System for Location-based Social Networks written by Rifeng Ding. This book was released on 2015. Available in PDF, EPUB and Kindle. Book excerpt: "In recent years, location-based social networks (LBSNs) has become more and more popular. As one of the key service in LBSNs, the location recommendation system has drawn much of attention from both industry and academia. According to existing work, link analysis-based methods have been proved to be effective inlocation recommendations for LBSNs. However, most of link analysis-based methods either overlook or overemphasize users' preferences. Recommendation systems that overlook users' preferences can only provide generic recommendation, while systems that overemphasize users' preference cannot recommend local popular locations that do not fit users' historical preferences. To address these issues, in this thesis, I propose a community-based location recommendation system, which takes both users' preferences and locations' popularity into account. Our system groups locations within the user-specified region into communities. Each community represents one location category and will generate a certain number of recommendations. More specifically, communities that represent user-favored categories and communities that contain large number of popular locations have higher priorities to recommend more locations. Besides, the number of recommendations of each community is dynamically calculated for different users at different regions. Thus, our system can cover both user-favored and local popular locations in its recommendations. In the evaluation, we acquire data from Foursquare, which contains 398,819 tips generated by 49,027 users who has visited the New York City. Our recommendation system outperforms the baseline approach with the precision and recall of 52.13%. and 80.01% respectively. The experimental result demonstrates that our system can provide more accurate recommendations with acceptable computation time for various types of users and solve the new-user problem as well." --