导　　师： 杨建梅

学科专业： 1201

授予学位： 博士

作　　者： ;

机构地区： 华南理工大学

摘　　要： 在没有市场机制引导的大众生产领域也能实现知识的分工与合作已经是一个不争的事实，但这种新兴生产方式下的产品的知识结构及演化机制却未被充分认识。本文以著名的开源软件——R软件为对象，研究其知识结构和开发者合作结构及其演化机制，以期达到认识开源软件乃至大众生产过程中知识聚集过程的目的。主要工作及结论如下： 首先研究体现R软件知识结构的软件包网络，发现随着时间的变化，网络规模呈幂律增长趋势，稠密度随规模增加而下降，网络凝聚性随规模增加而减弱，表明R软件中的知识呈加速凝聚的态势，知识结构呈现扩张趋势；同时，R软件的发展方式表现为最大连通片主体不断吸收聚集游离小连通片；在知识结构的扩展中，R软件保持了功能主轴不分叉的特性。对R软件的知识结构建立树关系网络模型分析，发现其存在核心结构，并成功识别出知识结构的功能社团。 进一步分析与知识结构形成紧密相关的开发者合作结构，发现它不但具有自身的演化特点，而且与R软件的知识结构存在耦合现象。这一发现揭示了从合作-技术双维度的新视角对开源软件进行分类研究的意义。通过对合作结构最大连通图的深入分析，发现合作结构有两种基本模式，一是以开发者为中心，二是以软件包为中心。 随后基于随机行动者导向模型，本文研究了R软件知识结构和开发者合作结构的内生演化机制以及两种结构的共演机制，发现了统计上显著的网络微观结构是驱动两个网络演化以及共演的内生机制。实证研究表明R软件包网络的演化中根节点及其子节点共同衍生新节点的机制在统计上是显著正面作用的；合作结构二模网演化中合作延续性机制是有显著的正面作用的；而软件包之间的技术联系对合作网络中的软件包共开发者关系形成有显著的正面促进作用。这些发现拓宽了对于网络演化机制的认识。 本文的主要贡献体现在：在知识分工新视角下，以R软件为代表，运用多种网络分析方法，尤其是动态演化的网络分析方法，研究大众生产知识聚集过程中的知识结构以及与之紧密相关的合作结构。在研究方法上，本文提出了基于无圈有向网络的树关系网络建模方法、能够同时识别核心与社团结构的Hub-社团分析方法，推广了随机行动者导向模型，提出了合作-技术这个物理-事理的分析框架。在研究内容上，本文深入剖析了R软件的知识结构、开发者合作结构，以及它们的内生演化机制。研究结论对认识大众生产模式下的知识产品生产过程具有理论价值，同时对软件开发过程的知识管理具有实践参考价值。 It is an indisputable fact that the knowledge division and cooperation can realizein the production fields of peer production. But the knowledge accumulation processunder this new mode of production has not been fully understand this new productionmode of. In this paper, the machanism and order behind OSS is explored from theperspect of the knowledge structure and the developer cooperation structure and itsevolution based on the famous R software as the research object. Main works andconclusions are as follows: First the study of the knowledge structure of R software shows: the networkscale submit to power law distirbution over time, as the scale growing, the density andthe cohesion decreases,which means the accelerated cohesive trend of the knowledgeof R software and the expansive trend of R software knowledge structure; R softwaredeveloped through that the maximal connected component constantly absorb andgather free and small connective components.As the expansion of knowledgestructure, the spindle of R software knowledge structure does not fork. Based on thetree relationship network model analysis, we found the existence of the core structurein the R software and successfully identify the functional communities in theknowledge structure. There exist a close relationship between knowledge structure and cooperativestructure of R software. The study shows there exist evolution properties in thecooperational structure and the study also found the relationship couple phenomenabetween co-developer packages networks and R package network, reflecting therelationship overlap at the dimensions of people and technolgy that inspiring a'technology-cooperation' perspective to explore the properties of OSS. Employing the stochastic actor oriented model/(SAOM/), the paper studied on theevolution machanism of knowledge structure and developers cooperational structureof R software. And the results reveals that there are some statistically significantendogenous network structure which drive these two structure evolution. Besidespreferential attachemnt which also know as indegree popularity, such as transitivity effect, four-cycles, out-indegree assortativity are statistical significant also. Thesefindings broaden the knowledge of the mechanisms for network evolution. The main contribution of this paper is the research of knowledge structure andcooperative structure of R software at a new perspective on the division of knowledge,with method of dynamic evolution complex network analysis. At the methodsperspective, this paper presents tree modeling method based on acyclic directednetwork and no-hub-community detection method which can indentify communityfrom the network with core and community both; generalizes SAOM to moreapplication fields; proposes a 'technology-cooperation' framework to analyse OSS.In conclusion, the research is helpful theoretically to understand production process ofpeer production and useful to the knowledge management of software development.

关 键 词： 软件 开源软件 知识结构 合作结构 随机行动者导向模型

分 类 号： [TP311.5]

领　　域： [自动化与计算机技术] [自动化与计算机技术]