Maximization Of Competitive And Cooperative Influence Spread In Social Networks

With the popularity of online social networks, the influence propagation of social networks has attracted much more attention. Social network influence maximization is the key problem of social network influence propagation. The influence maximization means that in a given social network with assigned diffusion model and assigned sets of seeds, finding out the specific set of seeds through which to maximize the influence spread. Social network influence maximization has applications in viral marketing, information diffusion, etc. In the fields of viral marketing and information spread, there exists the propagation of single information in social networks, and also exists competitive influence spread and cooperative influence spread. This thesis mainly focuses on the maximization of competitive and cooperative influence spread in social networks. Experiment results verify the feasibility and effectiveness of the methods in this thesis.The main contributions and novelties of this thesis are summarized as follows:? Studies parallely select seed set to maximize the influence spread in social networks. Based on k-shell decomposition, candidate shells generation, and heat diffusion model, this thesis proposes the candidate shells influence maximization (CSIM) as an effective method of selecting seeds in social networks.This study starts from improving the speed of selecting seeds. It can select seeds from candidate shells in parallel based on the method of candidate shells. Based on the heat diffusion model, introducing the time parameter, it can well simulate the influence spread in virus marketing.? Studies maximizing the spread of competitive influence in a social network oriented to viral marketing. This thesis proposes a method to maximize the competitive influence spread based on the extended linear threshold model, the analytical framework of submodularity, and the basic idea of greedy algorithm.This study starts from the real-world application of social network. To better reflect the realistic competitive influence spread in viral marketing, we propose the extended linear threshold mode (ELTM) to simulate the competitive influence spread. We further employ the greedy algorithm to approximately select seeds based on the analytical framework of submodular function.? Studies maximizing the spread of competitive influence in a social network oriented to information spread. This thesis proposes a method to select seeds for competitive influence spread maximization in social networks based on the possible graphs, diffusion model, the submodularity framework, and the improved greedy algorithm.This study starts from the possible graphs, which can resolve the hardness of computing the influence spared and improve the speed of selecting seeds. We employ the competitive influence spread model (CISM) to simulate the competitive influence process in possible graph. Further, we employ the cost-effective lazy forward (CELF) to approximately select seeds based on the analytical framework of submodularity.? Studies maximizing the spread of cooperative influence in a social network oriented to viral marketing. To simulate the cooperative influence spread in viral marketing, this thesis proposes the improved greedy algorithm to maximize the cooperative influence spread, based on the similarity model (SM), cooperative influence spread graph, the independent cascade model with accepted probability (ICMAP), and improved greedy algorithm.This study starts from the maximizing the cooperative influence spread. The cooperative influence spread graph of associated products can be obtained by the single influence spread graphs. To better reflect the products adoption, the ICMAP is proposed to simulate the cooperative influence spread in a social network. By directly estimating the influence of users with time range, the massive Monte Carlo computation can be avoided.