Host Profit Maximization For Competitive Viral Marketing In Billion-Scale Networks

Authors:
Yuqing Zhu California State University, Los Angeles, USA
Deying Li Renmin University of China, P.R. China

Abstract:

We study the problem to maximize the profit for a social network host who offers viral marketing to multiple company campaigners. Each campaigner has its special interest for social network users, and she pays the host commission when a target user adopts her product. The campaigners decide the cost they are willing to pay for viral marketing, and the host collects cost from all campaigners and uses it for marketing. We call our optimization problem Competitive PROfit maximization for the host (CPro), and its solution is the seed allocation for campaigners, such that the profit (commission) campaigners given back to the host is maximized. CPro is NP-hard with a non-monotone and non-submodular objective function, which means existing techniques in influence or profit maximization cannot give guaranteed performance. To solve this issue, we design an efficient approximation algorithm that works on billion-scale networks, and more importantly, we give the performance bound of our algorithm. As far as we know, this is the first bounded scalable approximation algorithm for competitive profit maximization. A comprehensive set of experiments are set on various real networks with up to several billion edges from diverse disciplines, and our solution identifies the top choices for the host in only a few minutes on network that contains 1.5 billion edges.

You may want to know: