Detecting information source in diffusion networks

With the rapid development of the Internet, information can be widely diffused through the social network. It is critical to locate the information source under some circumstances in nowadays growing network. In this paper, we propose an effective algorithm to locate the information source. First, we apply Fiedler vector to partition the network into several node clusters, where observers are selected. Then a Maximum-Likelihood source estimator is used to locate the source according to the information offered by observers. We take the full network topology into account instead of the local influence of some nodes selected randomly or those based on degree of centrality in the network. Observers chosen in node clusters achieve higher accuracy rate than those selected in whole networks under the circumstance of the same percentage of the observers selected.