Research On Node Importance Ranking Methods Based On Network Entropy

In a network,nodes of different importance have different effects on network performance,and the more important nodes have a greater impact on the network.For example,in a transportation network,an accident on an important node may cause congestion of the entire transportation network;in a communication network,if the core node is damaged,the throughput or security of the network will also drop sharply.Therefore,it is necessary to study the importance of different nodes in the network.When the network is under attack,if the current defense capability cannot protect all nodes,it is a reasonable choice to prioritize the protection of the key nodes in the network based on the importance ranking.Generally,the importance of a network node is not only related to the topology of the network,but also related to the current network load status.When the load status of the network changes,the importance of the node also changes.However,most of the currently known node importance ranking methods are based on static analysis of the network topology.When the network load is unbalanced,these methods generally fail to reflect the change in the importance of nodes with changes in network load.In response to the above problems,this thesis proposes two node importance ranking methods that combine network topology and network load changes: 1)Entropy Differences(En D)method based on network entropy;2)Mix R(Mix Ranking)method based on rate of change of network entropy changes(REC).Among them,the basic theory of the En D algorithm : deleting a node in the network will cause changes in the network topology and network load.The more important the deleted node,the greater the change in the corresponding network entropy.Therefore,the change in network entropy can be used to measure the importance of the deleted node from one side;the basic principle of the Mix R method: by measuring how fast the network entropy changes near a node,the nodes with large fluctuations in network traffic can be identified,and the importance of nodes can also be evaluated in combination with the network topology.The two methods proposed in this thesis both combine network topology and network dynamic load changes.They both are typical evaluation methods of node importance that combine static and dynamic factors.The experimental results show that it is different from the known similar methods,such as the degree centrality method(Degree Centrality,DC),the closeness centrality method(Closeness Centrality,CC),the eigenvalue centrality method(Eigenvalue Centrality,EC)and the semi-Compared with Semi-local Centrality.According to the dynamic changes of network load,the two algorithms proposed in this thesis have higher efficiency and accuracy in ranking the importance of nodes.