Research And Realization Of Propagation Model And Influence Maximization Algorithm Of Innovation Behavior

In the marketing field,while promoting a product,a company asks some influential individuals to test the free samples so that the product can be widely known with the strategies of “Word-of mouth”.The key problem of this marking strategy is that how to spend limited budget to find the most influential individuals who can eventually influence the largest number of people.This is the initial motivation of the influence maximization problem.This problem has become a focus in academia since it was introduced to the field of social network,and scholars have proposed different models and algorithms to resolve it.Firstly,this thesis introduces the relevant theoretical knowledge of the influence maximization problem,especially the Independent Cascade(IC for short)and the Liner Threshold(LT for short)propagation models.Then this thesis explores the advantages and disadvantages of exiting algorithms in detail.The exiting propagation models are not realistic because they ignore the complexity of human.This thesis proposes a social learning model in which a participant learns the strategy of his neighbor with the biggest benefit to make it correspond to reality more closely.This thesis presents a mixed model,combining social learning model and LT model,and proved by experiments that the mixed model has better effect than the single model.Then this thesis proposes a new algorithm based on social learning model.This new algorithm takes the payoff of each node into account and avoids the centralization of the initial active node to make the new activity spread widely,and experiments prove that this algorithm has better effect than the HighDegree and Random algorithm.