基于科技引用网络结构算法的科学计量新方法研究

发布时间：2018-03-29 05:22

本文选题：TSU引用网络　切入点：引用代　出处：《浙江大学》2012年博士论文

【摘要】：研究评价及其计量指标是科学学与科学技术管理过程中的重要决策工具。本文围绕目前科学计量学领域面临的科学进化过程中的复杂网络与评价指标的“平面化”、科学与技术的飞速发展与评价指标的“重论文、轻专利”之间的矛盾与挑战,以及引文分析存在的对专利计量关注不够、缺乏结构关联度指标、对参考文献的计量研究偏少等薄弱点与亟待解决的问题,进行了基于引用网络结构算法的科学计量新方法研究。本研究将图论与引用网络分析相结合,主要涵盖了以下一些问题：(1)基于"TSU-CN+E-CN"结构观的系列新指标的理论定义；(2)新指标的一般应用；(3)基于TSU-CN而展开的多个引用代科学计量方法的机理研究及其深层研究的具体应用；(4)新指标在专利引用网络中的延伸发展及其具体应用。本论文的研究内容和研究成果主要有以下几个方面：第一,在图论基础上,创造性地提出了基于引用网络“目标节点-子网-全网”的TSU-CN结构分析框架。将TSU-CN与向心引用网络(E-CN)相结合,可以合理解析科学研究基于前人积累而进化的特点以及科学研究中“研究个体-小科学-大科学”的相对特性。经过本研究大量的理论研究与现实数据分析表明,在"TSU-CN+E-CN"结构框架下,可以解析引用代影响力的分析机理、文献结构指标的算法机理、专利结构指标的算法机理。因此,"TSU-CN+E-CN"结构分析法有望为引用网络的科学计量提供新的分析框架。第二,本研究针对目前涉及论文间接影响力的算法中存在的没有考虑网络的方向性和节点的重复性问题,引入集合与多重集合理论,将图论、数学、信息分析理论与技术、复杂网络分析技术、可视化分析技术有效结合并加以改进,首次提出了科学引用代结构概念及代结构的几个分型Gs,Gm,Hs,Hm,并应用指数母函数算法,成功解决了科技引用网络中前后三代影响力的算法问题(同时考虑直接影响和间接影响),在研究方法和分析技术上突破了国际科学界科学家影响力计算只考虑直接引用的现状,包括总引用代指数、跨代指数、跨代h指数、总影响指标、引用代总h指数(Th)、引用代总影响力(TI)、反向引用代总聚合度(CG)的系统定义及精确算法。为准确评价论文的学术贡献提供了一套工具。第三,在假设专利“可视技术价值”与“利润”呈正相关的基础上,应用"TSU-CN+E-CN"结构分析原理,创造性地将专利技术价值评价与目标专利在动态专利引用网络中的专利子网结构中的“表现”相对接,并将现实世界专利技术价值测量与专利子网结构特性分析相联系,对Rousseau教授的论文向心引用网络结构指标,进行了在“专利环境”下的进一步改良和OI(P)、RR (P)、CC(P)、RC (P)的定义。实证研究结果表明,本研究设计的结构指标在表征专利技术价值上有显著意义。经计算机大样本模拟分析,验证了“实证研究”数据与结果的可靠性和稳定性。第四,基于引文分析的应用需求,本研究首次提出“结构研究应该成为引文分析三大基本应用功能之一”,首次将新型的引用网络结构指标外部生长指数与时间序列分析技术结合应用,根据不同时间维度提出外部生长指数的年序堆积序列、同步序列、同步累积序列、矩阵型序列的理论定义及算法。根据现实世界中的真实样例观察,上述不同维度的时间序列能从不同角度映射论文影响力生长的内部特征,不仅适合于评估实践中对单篇论文的评价,也适合于对一组论文集合进行多角度评价。本论文研究结果在基于引用网络结构算法的科学计量理论方面提出了系列新概念和新指标,包括基于引用网络的TSU-CN结构分析框架；科学引用代结构概念及代结构的几个分型Gs, Gm, Hs, Hm,引用网络前后三代影响力的系列算法与定义；引用网络结构指标的概念和理论算法,用于动态引用网络分析的四种时间序列分析模型；提出并验证了用专利子网结构指标评估某专利技术价值的假设与理论。同时,本研究建立的在引文分析和专利技术价值的创新分析方法,可以为研究评价实践中提供指导和方法工具。
[Abstract]:The research and evaluation and its measurement indexes are important decision - making tools in the process of science and technology management . This paper focuses on the contradictions and challenges between the " planarization " of complex networks and evaluation indexes in the process of scientific evolution in the field of scientific metrology , the contradictions and challenges between the rapid development of science and technology and the " heavy paper and light patent " of the evaluation indexes , as well as the problems of insufficient attention to patent measurement , lack of structural correlation degree indexes , and the like of references , and the scientific measurement new method based on the reference network structure algorithm .
This study combines graph theory with reference network analysis , which mainly covers some of the following problems : ( 1 ) the theoretical definition of the series of new indicators based on " TSU - CN + E - CN " structure view ;
( 2 ) General application of new index ;
( 3 ) Based on TSU - CN , the mechanism research and concrete application of several reference generation methods are studied .
( 4 ) The extension development of the new index in the patent reference network and its application . The research contents and the research results of this paper mainly include the following aspects :
First , on the basis of graph theory , the TSU - CN structure analysis framework based on the reference network " target node - subnet - global network " is creatively put forward . By combining the TSU - CN and the centripetal reference network ( E - CN ) , it is possible to analyze the analysis mechanism of the reference generation influence , the algorithm mechanism of the document structure index and the algorithm mechanism of the patent structure index in the framework of the " TSU - CN + E - CN " structure . Therefore , the " TSU - CN + E - CN " structural analysis method is expected to provide a new analysis framework for the scientific measurement of the referenced network .
Secondly , aiming at the problems of network directionality and node ' s repeatability existing in the algorithm involving the indirect influence of the paper , we introduce the theory of graph theory , mathematics , information analysis theory and technique , complex network analysis technology and visual analysis technology .
Third , on the basis of the positive correlation between the " visible technology value " and " profit " of the patent , the paper applies the " TSU - CN + E - CN " structure analysis principle , and creatively combines the value evaluation of the patent technology with the " performance " in the patent sub - network structure in the dynamic patent reference network .
Fourthly , based on the application requirement of citation analysis , the first proposed structure research should be one of the three basic application functions of citation analysis . The new reference network structure index external growth index and time series analysis technique are firstly used in combination with the application . According to the real example in the real world , the time series of the above different dimensions can map the internal characteristics of the influence growth of the paper from different angles , which is not only suitable for the evaluation of single articles in the evaluation practice , but also for the multi - angle evaluation of a set of paper collections .
The results of this paper put forward a series of new concepts and new indexes in the theory of scientific measurement based on the reference network structure algorithm , including the framework of TSU - CN structure analysis based on the reference network ;
A series of algorithms and definitions of the influence of three generations before and after the network are referenced by the concepts of structure and structure .
Based on the concept and theory algorithm of network structure index , four time series analysis models for dynamic reference network analysis are introduced .
This paper puts forward and verified the hypothesis and theory of using patent net structure index to evaluate the value of a patent technology . At the same time , this study establishes an innovative analytical method in citation analysis and patent technology value , which can provide guidance and method tools for the research and evaluation practice .

【学位授予单位】：浙江大学
【学位级别】：博士
【学位授予年份】：2012
【分类号】：G647

【参考文献】