Date of Award
5-2015
Degree Type
Thesis
Degree Name
Master of Science in Computer Science (MSCS)
Department
Computer Science
First Advisor
Dr. Jing (Selena) He
Abstract
Nowadays, social networks are considered as the very important medium for the spreading of information, innovations, ideas and influences among individuals. Viral marketing is a most prominent marketing strategy using word-of-mouth advertising in social networks. The key problem with the viral marketing is to find the set of influential users or seeds, who, when convinced to adopt an innovation or idea, shall influence other users in the network, leading to large number of adoptions. In our study, we propose and study the competitive viral marketing problem from the host perspective, where the host of the social network sells the viral marketing campaigns to its customers and keeps control of the allocation of seeds. Seeds are allocated in such a way that it creates the bang for the buck for each company. We propose a new diffusion model considering both negative and positive influences. Moreover, we propose a novel problem, named Blocking Negative Influential Node Set (BNINS) selection problem, to identify the positive node set such that the number of negatively activated nodes is minimized for all competitors. Then we proposed a solution to the BNINS problem and conducted simulations to validate the proposed solution. We also compare our work with the related work to check the performance of BNINS-GREEDY under different metrics and we observed that BNINS-GREEDY outperforms the others' algorithm. For Random Graph, on average, BNINS-GREEDY blocks the negative influence 17.22% more than CLDAG. At the same time, it achieves 7.6% more positive influence propagation than CLDAG.