Proposing a Method for Ranking Nodes in Complex Networks

Please enable javascript in your browser.

Esnaashari, Marzieh | 2019

546 Viewed

Type of Document: M.Sc. Thesis
Language: Farsi
Document No: 52514 (01)
University: Sharif University of Technology
Department: Industrial Engineering
Advisor(s): Mahlooji, Hashem; Safaei Semnani, Farshad
Abstract:
A distinct viewpoint is adopted by each centrality to analyze a network and rank its nodes. This study aims to introduce a novel centrality that ranks the nodes of a network more effectively. In this respect, a function of five centralities, namely betweenness, closeness, agent vector, degree, and Katz, is introduced to maximize the connected components of the network after ranking its nodes and deleting the first twenty ones. The proposed centrality functions better than the other mentioned centralities. Among the networks simulated to evaluate the centrality, it functions better in Erdos-Renyi and small-world networks, both of whom being based on the Poisson degree distribution, and also scale-free networks with Parto distribution. Other than simulated networks, four real ones are utilized to ensure the precise performance of the proposed centrality. All the chosen networks are weightless and undirected. The results display that the merged centrality of this study performs superiorly than the other centralities. The distribution degree of these networks are detected to be similar to that of Poisson
Keywords:
Influential Nodes ; Complex Network ; Network Centrality ; Particles Swarm Optimization (PSO) ; Nodes Ranking

No TOC