Abstract
Location-Based Social Networks (LBSN) are a vast source for personalized geo-social user data with more and more users adding spatial information to their posts and tweets. The potential of spatially rich user models lies in the aggregation of different networks with each network adding a piece to the digital footprint of a user. These user models in turn allow new ways of analysis and let users benefit through novel applications. We describe an aggregation approach that extracts data from Facebook, Google+, Twitter, and Foursquare and generates a materialized spatial schema. For rapid prototyping, Preference SQL is applied to generate extensive user models from aggregated data with minimal modeling effort. In the process, explicit user information is augmented with implicit preference information. Finally, similarity analysis based on these user models is defined and applied to a promising application scenario of user-to-user recommendation. To the best of our knowledge, we are the first to present an integrated framework for the synthesis, modeling, and analysis of multi-source LBSN data.
Authors
Florian Wenzel, Werner Kießling
Publication
Proceedings of the 1st Int. workshop on managing and mining enriched geo-spatial data (GeoRich) in conjunction with SIGMOD, 27th June, 2014, Snowbird, Utah, USA