Discussions And Experiments For Algorithms Of Influence Maximization On Social Network

The arrival of internet and big data at 21st century have made such significant impact that the activity of information transmitting has reached an unprecedented level. Among various internet applications, social network and social network marketing has become a hot topic for research.A common problem for social network marketing is how to define the scope for marketing and how to maximize the impact for marketing campaigns. For a closely connected group of people, social network can be viewed as a graph structure, with individuals as vertices, and activities between people as edge weights. Thus we get a social network for people connections.With graph of social network defined, we can design and implement corresponding algorithm targeting to maximize the marketing impact. Classic algorithms such as CELF Greedy, MIA and DegreeDiscount are well developed in the area, while each has its original trait and advantage.This paper implements these algorithms and other frequently used algorithms (PageRank, HITS). Using real world data, the best strategy for influence maximization on social network is discussed. It is shown that CELF Greedy can maintain a steady output, while algorithms such as MIA and DegreeDiscount needs great effort on parameter settings, which requires multiple trails and adjustments.As algorithm designer for a real world project the author implements the algorithms in a real world project.As part of the contribution for this paper, a new approach for evaluating influence maximization problem is proposed. This approach can significantly reduce running time for current algorithms.