Recommendation in location based social networks

Abstract

Recommendation in Location Based Social Networks (LBSNs) is an emerging research topic. Compared to conventional recommendation systems, there is much more information could be utilized in a LBSN recommendation system. In this report, the author first proposes a new method of utilizing spatial sensitivity to generate dynamic weightage average score for combining the User Collaborative Filtering and the Geographical Influence in a personalized fashion for each single user. Temporal information in check-ins which interconnects users and point-of-interests have significant value in LBSNs. In this report, the author also discusses how to leverage temporal information in the User Collaborative Filtering and the Geographical Influence. Finally, a unified framework for all proposed methods is defined in this report. A detailed empirical experiment is included in this report which provides a comprehensive comparison of the performance of proposed methods against baseline methods.