Research On Identifying Discoverers In Co-authorship Networks And Its Predictability

Network science has been firmly established as a framework for studying complex system and analyzing large and complex data these days.The vital nodes of complex networks refer to those who play more important roles in the organization of network structure or the dynamic behaviours of networked systems.And identifying vital nodes in complex networks has always been a hot research topic in network science.Vital nodes can not only be utilized to understand network structure and function,but also applied into different practical fields including network attack,synchronization,and so on.In social networked systems,there are some special vital nodes,which are known as discoverers,that have specific interaction with other nodes and have the ability to predict success.In fact,discoverers are ubiquitous in networked systems including e-commerce platforms and social networks.And in this paper,we found discoverers in the co-authorship network.The previous study has shown that junior researchers can benefit a lot from the collabora-tion with top scientists in co-authorship networks.However,we find that collaborations with discoverers still lead to a similar or even more powerful competitive advantage and predictive power.Therefore,discoverers identification is of significance in both theory studying and practical application.The thesis consists of following main parts:1.We first introduce some basic models and definition of network science,vital nodes identification,and science of science.Then the development of discoverers identification problem in complex networks is reviewed.2.We propose a universal vital nodes identification method for simple time-independent networks by Achlioptas process.The excellent performance of the method is verified by different experiments including the influence of network structure(network density,model,clustering and assortativity)and non-structure factors(protection strategy and noise).And we further analyze a practical advantage of the new method named multi-solution and the computation complexity of the new method.3.We demonstrate that different researchers in the co-authorship network have differ-ent discovery abilities,and a model is designed to identify discoverers in the co-authorship network.Then we study the relationship between discoverers and junior researchers,which leads to a promising result that collaboration with discoverers can prompt a junior researcher's career.4.We demonstrate that discoverers can offer success prediction for a junior researcher's career.And We study the predictive power of discoverers by comparing it with other cen-trality indices.