Abstract
Many phenomena can be model as epidemic in social network, such as viruses, rumors,ideologies, or marketing etc. When there exist two competitors who try to maximize each spread in the network, there comes the question that what nodes target in the very beginning to achieve the maximum spread. The epidemic model we use is called SI1I2(Susceptible-Infected1-Infected2) model, where there are two viruses spreading in the network. We fi nd that there are some problems when extending the approximation from one virus to two viruses. We derive the appropriate approximation according to community detection result. We compare several strategies for the initial chosen nodes for each competitor. Choosing top degree nodes in each community is the strategy we give which has more influence than others according to centrality properties. Our approximation agrees with these results.