Research On Vital Nodes Identification In Complex Network

In the real world, network systems can be seen everywhere, complex networks is aabstraction of complex network system drawn from the real world. Such as the Internet, socialnetworks, cooperative network of scientists, aerospace-line network, the epidemic spread ofnetworks, crime information network, routing network, etc. The complex network research isbecoming more and more important. Complex network is increasingly becoming a veryimportant research area.There are many methods to identify vital nodes; these methods are mostly based on thenodes, such as: degree centrality algorithm, betweenness centrality algorithm, closenesscentrality algorithm, sub-graph centrality algorithm, eigenvector centrality algorithm, andinformation centrality algorithm. However, these algorithms are focusing on a particularprospect to identify the vital nodes in the network, and even only concern the nodes’importance, thereby losing some features of the networks. So the accuracy and effectivenessof the vital nodes we identify is not very good.By reading Mining Vital Nodes in Complex Network and other related literature reviews.We know have a certain understanding of the aspects of the concept of complex networks,characteristics and development theory of the complex network vital nodes detection method.And we summarized the advantages and disadvantages of these vital nodes detection methods.In addition, this paper studies the new centrality (NC) algorithm in-depth. This method is anew method to detect important nodes these years, by calculating the edge clusteringcoefficient to assess the nodes’ importance. NC algorithm has some advantages than thetraditional node detection method. It can detect more important node in the preferred list, butalso it has shortcomings. The method doesn’t considerate the non-directly connected edges’contribution. Thereby lose the some other features and the accuracy and validity reducedIn this paper, we propose a method called Link Route Centrality measure based on theRoute Contribution Coefficient. The non-directly connected edges’ contribution has beenconsiderate. In order to compare the performance of LRC and NC methods, we selected thesix traditional methods (DC, BC, CC, SC, EC, and IC) as the reference. And we name theOverlapping nodes number as an indicator to compare the LRC and NC methods performance.In three real-world social networks, Zachary’s Karate Club network, Dolphin Social Networkand the American College Football Network tests we achieved better result than NC.Meanwhile, to further analyze the effectiveness of LRC method in the real world, we selectthe terrorist network, and through experimental comparison LRC are better than the NC inreal society. In small scale static networks, the vital nodes identified by Link Route Centralitymeasure have higher accuracy. But in the real world, the complex network has larger scaleand higher complexity. We will continue study and improve the vital nodes identify method tomake it better to be applied to the real network.