The Detection And Location Of The Information Source In Social Network

The rapid development of Internet technology has promoted the popularization and development of online social network. At the same time, the rich social network has provided the convenient medium and mode of transmission for the information dissemination. However, due to the huge amount of users in social networks, the expression of free expression of speech, which makes some of the bad information about individuals or businesses in the network in general, people's lives and production has a profound impact. If fast and accurate, it is of great practical significance to find the source of information dissemination, to control and narrow the scope of dissemination of bad information, which can greatly reduce its negative impact on people. So, the research of information source detection has an important practical significance.In the aspect of information source localization, By using the observational information of observers in the network, this paper constructs a maximum likelihood estimator in breadth-first traversal way which is based on the shortest path hypothesis in the process of information propagation. The candidate nodes in the network perform the estimation and select the node with the largest estimated value as the result of the information source location. In the part of the selection of observation points, the deployment of observing point is selected in several ways, such as High-Degree, High-Betweenness, High-Closeness and random selection. On ER random network and BA network data set,the results show that the proposed estimator has high reliability with ER network and BA network and this method is effective and reliable.When researching the problem of deploying strategy in observing point, we find that it only chooses some special nodes in the complex network and it does not consider the whole structure of the network completely.By analyzing, the higher the value of the shortest path from the observation point to the information source is, the higher the similarity between the theoretical propagation delay and the actual propagation delay of the observation point are, and at the same time, the higher the accuracy of the information source location is expected to be.So, the Laplacian matrix of network and its Fiedler vectors are taken into consideration and used to segment the network into multiple subgraphs where the observation points are deployed, and then the maximum likelihood estimation is used to locate the information source.Experiments show that the localization effect is superior to the traditional observation point deployment method in ER model network, BA model network and real network data.