Discovering the Most Influential Geo-Social Object Using Location Based Social Network Data

In the scope of knowledge engineering, discovering the most influential geo-social object is one of the most extensively studied problems, where the reverse top-k queries can be used as a key technique to detect the influence set, also refereed as potential customers in this paper. By issuing reverse top-k queries, merchants can get the knowledge of the potential influence of their products and then make effective decisions in business promotion applications. In this paper, we study the problem of discovering most influential geo-social object using LBSN data. More specifically, given a set $\mathcal{U}$ of LBSN users, a set $\mathcal{O}$ of geo-social objects, and a set $\mathcal{O}$ of candidate objects extracted from $\mathcal{O}_{c}$, we attempt to find the optimal one in $\mathcal{C}$ that has the largest potential influence, where the potential influence of an object is defined by the size of users in its reverse top-k query results. Such problem is practical for merchants to monitor which product among all the products is the most popular with the potential customers. A baseline approach based on a batch processing framework is proposed to facilitate answering this problem. On the top of this solution, a series of optimizations are integrated to further improve its performance and make it more efficient in practise. Experiments on two datasets are conducted to verify the effectiveness and efficiency of the proposed methods.