Location Recommendation for Mobile Users in Location-Based Social Networks

Open Access
Ference, Gregory David
Graduate Program:
Computer Science and Engineering
Master of Science
Document Type:
Master Thesis
Date of Defense:
March 26, 2013
Committee Members:
  • Wang Chien Lee, Thesis Advisor
  • Location Recommendation
  • Location-Based Social Networks
  • Collaborative Filtering
  • K-Nearest Diverse Neighbor
  • Spatial Diversity
Location-based services have become popular in the twenty-first century due to technological advances, such as mobile and online social networking. One of its key features is location recommendation, which encourages users to explore new locations. In this thesis, we propose methods for recommending locations that are interesting for the user in terms of closeness in proximity and spatial diversity in relation to the user's current location. First, most previous research on location recommendation services in location-based social networks (LBSNs) makes recommendations without considering where the targeted user is currently located. Such services, as a result, may recommend to a user traveling out of town a place close to her hometown. In this thesis, we study the issues in making location recommendations for out-of-town users by taking into account user preference, social influence and geographical proximity. Accordingly, we propose a collaborative recommendation framework, called User Preference, Proximity and Social-Based Collaborative Filtering (UPS-CF), to make location recommendation for mobile users in LBSNs. We validate our ideas by comprehensive experiments using real datasets collected from Foursquare and Gowalla. By comparing two baseline algorithms (i.e., popularity-based and distance-based approaches) and conventional collaborative filtering approach (and its variants), we show that UPS-CF exhibits the best performance. Additionally, we find that when users are in town, preference derived from similar users is important while social influence becomes more important as a user is out of town. Second, studies have shown that users prefer diversity in their recommendation results, but few works have considered spatial diversity in terms of recommending locations. In this thesis, we investigate the k-nearest diverse neighbor problem, which chooses locations that are spatially diverse as well as close in proximity to the user's current location. By fixing deficiencies in a state-of-the-art framework, we first propose the Modified Index-Based Diverse Browsing (Mod-IBDB) framework. In addition, we propose the Distance-Based Diverse Browsing (DBDB) framework that improves upon an initial solution to provide spatially diverse and nearby locations. We develop two versions based upon selecting the initial framework: Distance-First DBDB and Diversity-First DBDB. Using Foursquare and Gowalla datasets as well as synthetic datasets, we show that Mod-IBDB improves upon the performance of its predecessor. In addition, we show that DBDB methods outperform the mentioned state-of-the-art algorithms (as well as the k-nearest neighbor baseline) in terms of proximity and spatial diversity while maintaining high efficiency.