Ranking Node Influence According To Dynamic Network Load

The topology and network load change quickly in the networks.Concerning the network load changes,nodes at different locations have unequal influences with respect to the directions of packet transmissions.Some nodes are more influential,such as a network tower,endpoint device or a signal booster.Damaging these influential nodes will drastically decrease the mobile network's throughput and security,whereas the negative influence from an uninfluential node's breakdown is very limited.Therefore,if all the nodes can't be protected in a network,the nodes need to be protected based on their influences to achieve better network security and robustness.In general,network load is dynamic.Many methods have been proposed to evaluate the influence of nodes.However,most of the methods are applicable to undirected,unweighted and static topology network.There are few methods analyzing and ranking nodes importance according to network load and temporal changes in directed and dynamic network.In order to address the issues mentioned above,the thesis devised three dynamic algorithms--NodeRank,Direct Principal Component Ranking(DPCR)and Comprehensive Principal Component Ranking(CPCR)algorithms,to rank node influence according to dynamic network load changes.NodeRank only considers network linked weights and has higher efficiency;DPCR and CPCR consider the transmission direction of data packets and achieve higher accuracy.DPCR applies the rule that the most important node has the minimum value in the corresponding column of the cosine matrix which is calculated by Principal Component Analysis(PCA)and cosine similarity;CPCR applies a moderate rule that uses linear-weight comprehensive evaluation which takes the importance from each principal component in the measurement.To the best of our knowledge,DPCR and CPCR are the first known schemes that can rank node influence with respect to the directions of packet transmissions in the networks.Compared to the known methods,such as Degree Centrality(DC),Betweenness Centrality(BC),Closeness Centrality(CC)and Eigenvalue Centrality(EC),in the simulation experiment,the three proposed algorithms have better efficiency and accuracy to rank the node influence in the dynamic network.