返回文章

Figure 1.  The proposed competitive linear threshold model with state transition.

Figure 2.  A simple network with ten nodes for illustrating the CLTST model.

Figure 3.  An example of community detection.

Figure 4.  The spreading influence against time t on nine datasets obtained by CRB and six comparision algorithms, where the spreading influence represents the number of activated nodes at time t in the range between 0 and 25. The ratio of rumor and anti-rumor seeds is fixed as one percent of the size of datasets. The F(t) values are obtained under the CLTST model by averaging 200 independent runs.

Figure 5.  The stable propagation influence of seven algorithms on nine datasets. The ratio of rumor is fixed as one percent of the size of datasets, and the ratio of anti-rumor seeds varies in a certain interval (|S_A| ∈ [0.001,0.01] × N) to better depict the performance variation. The stable propagation influence values are obtained under the CLTST model by averaging 200 independent runs.

Figure 6.  The stable propagation influence of seven algorithms on nine datasets. The ratio of rumor seeds varies in a certain interval (|S_R| ∈ [0.001,0.01] × N) to better depict the performance variation, and the ratio of rumor seeds is fixed as one percent of the size of datasets. The stable propagation influence values are obtained under the CLTST model by averaging 200 independent runs.

Figure 7.  The average distance d_mean (under ten different anti-rumor seed ratios p ∈ [0.001,0.01]) between each pair of anti-rumor seeds obtained by different algorithms on nine real-world datasets.