| Summary: | 博士 === 國立交通大學 === 資訊科學與工程研究所 === 106 === The selection of influential network spreaders can be used to assist information diffusion on an online social website, to make a strategy for viral marketing, to monitor and control epidemic disease outbreak in a population, to analyze cascading failures in electrical power grids and the Internet, among many others. In this dissertation, we proposed solutions of an influence measure, a community detection algorithm and a community-centroid measure for selecting the maximal or multiple influential network spreaders. The computer simulation was used to verify the network-spreading ability of selected network spreaders. In our first research work, we proposed a two-step framework as a robust and reliable influence measure, which combines global diversity and local feature for selecting the maximal influential network spreader. The experiment results indicate that our proposed influence measure performs well and insensitively in network-spreading simulation for various network datasets. In our second research work, we proposed a rule-based hierarchical arc-merging (HAM) algorithm, which has good performance efficiency for community detection tasks of large-scale real-world networks; we also proposed a community-centroid measure to select the central nodes of community structures as multiple influential network spreaders. The experiment results indicate that our measure could result in the maximal 40% (or an average 30%) benefit in network-spreading simulations. In final, according to our results, we provided a decision diagram for determining an appropriate selection strategy of influential network spreaders for a network.
|
|---|
