VRer: Context-Based Venue Recommendation using embedded space ranking SVM in location-based social network

Abstract

Venue recommendation has attracted a lot of research attention with the rapid development of Location-Based Social Networks. The effectiveness of venue recommendation largely depends on how well it captures users’ contexts or preferences. However, it is quite difficult, if not impossible, to capture the whole information about users’ preferences. In addition, users’ preferences are often heterogeneous (i.e., some preferences are static and common to all users while some preferences are dynamic and diverse). Existing venue recommendation does not well address the aforementioned issues and often recommends the most popular, the cheapest, or the closest venues based on simple contexts.In this paper, we cast the venue recommendation as a ranking problem and propose a recommendation framework named VRer (Context-Based Venue Recommendation using embedded space ranking SVM) employing an embedded space ranking SVM model to separate the venues in terms of different characteristics. Our proposed approach makes use of ‘check-in’ data to capture users’ preferences and utilizes a machine learning model to tune the importance of different factors in ranking. The major contribution of this paper are: (1) VRer combines various contexts (e.g., the temporal influence and the category of locations) with the check-in records to capture individual heterogeneous preferences; (2) we propose an embedded space ranking SVM optimizing the learning function to reduce the time consumption of training the personalized recommendation model for each group or user; (3) we evaluate our proposed approach against a real world LBSN and compare it with other baseline methods. Experimental results demonstrate the benefits of our proposed approach.