A Study Of User Recommendation Algorithms Based On Social Influence Analysis

With the widespread use of internet, more and more people could share their ideas on online social networks (such as Facebook, Twitter), and social networks have playing an important role in information diffusions. Recent years have witnessed the increasing interest in exploiting social influence in social networks for many applications, such as viral marketing.Most existing research has focused on identifying a subset of influential individuals with the maximum influence spread. However, in the real-world scenarios, many individuals also care about the influence of herself and want to improve it. Hence, in this paper, we first consider such a problem that maximizing a target individual’s influence by recommending new social users. In other words, if a given individual/node makes new links with our recommended users then she will get the maximum influence gain. Additionally, most related works about the product marketing have focused on the influence between users, and paid less attention to the influence between companies and users. However, social users are usually influenced simultaneously by multiple companies, and both these social influences and the user interest will contribute to the user consumption decisions. Hence, we analysis the influece between companies and users and study the problem of the multiple influence-based potential user recommendation. In the end, the main contributions of this paper can be summarized as follows:(1) Research on the problem of maximizing a target individual’s influence by recommending new social users. We formulate this problem as an optimization problem and propose the corresponding objective function. As it is intractable to obtain the optimal solution, we propose greedy algorithms with a performance guarantee by exploiting the submodular property. Furthermore, we study the optimization problem under a specific influence propagation model (i.e., Linear model) and propose a much faster algorithm (uBound), which can handle large scale networks without sacrificing accuracy. Finally, the experimental results validate the effectiveness and efficiency of our proposed algorithms.(2) Research on the problem of the multiple influence-based potential user recommendation. We first formulate this problem as an Identifying Hesitant and Interested Customers (IHIC) problem, where we argue that these valuable users should have the best balanced influence entropy (being "Hesitant") and utility scores (being "Interested"). Then, we design a novel framework and propose specific algorithms to solve this problem. Finally, extensive experiments on two real-world datasets validate the effectiveness and the efficiency of our proposed approaches.