Location Based Scalable Recommendations Using Social Relationships

Abstract

Over the last few years, Location Based Services (LBS) has gained huge popularity due to the tremendous advancement of mobile handheld devices. Some of the applications of LBS include recommending nearby restaurants or shopping malls, local news, concerts or events, and local advertisements offering prizes or discounts. These applications are dependent on the location of the users. In this paper, we propose a location based recommendation framework that not only considers the location of the users but also utilizes their social connections in order to improve the recommendation accuracy and scalability. The proposed work recommends restaurants to users using Collaborative Filtering (CF) techniques and while recommending to a target user, it will consider only the ratings of those users who are also her friends in a social network. For this, we compute the similarity among the target user and her friends. Typically the most costly step of any CF algorithm is the similarity computation and since we are calculating similarity only with the friends of a target user bypassing the entire user set, this should definitely enhance the scalability of the system. In this work, we further classify the friendship relations present in social network into immediate friends and distant friends. Experimental evaluations using Foursquare dataset verify that better recommendation accuracy can be achieved considering the opinions of the friends rather than considering the entire set of users.