IT行业经济与管理指标测度与预报模型实证研究

【作者】 曹济；

【导师】 刘黎明；

【作者基本信息】 首都经济贸易大学 ， 数量经济， 2010， 博士

【摘要】 为了解决我国IT行业项目管理所面临的工期延误、费用超支、质量差强人意等典型问题,本文进行分析后认为造成这些问题的根本原因之一在于缺乏为IT行业项目设置量化目标的客观方法,因而本文的主要目标在于为IT行业项目建立关键管理指标体系,并研究这些指标之间的定量依赖关系。论文主要设计了IT行业项目质量和管理关键指标体系,并采用定量分析方法对指标之间的关系进行了分析和研究。对于IT行业项目质量和管理指标体系及其关系研究论文采用了定性方法和定量方法相结合的研究方法。定性分析方法主要用于建立指标体系,通过查阅IT行业项目管理领域以及相关领域的参考文献,并结合作者在IT行业项目量化管理领域的咨询实践,对业界三种主要的IT行业项目管理相关模型——PMBOK、ICB以及CMMI模型进行了分析,认为当前的项目管理模型缺乏关于对建立IT行业项目质量和管理指标体系的指导。在项目管理模型分析的基础上,又对业界所涉及常见指标体系——ISO15939、SEI indicator ISBSG data questionnaires、SPR indicator和Bangalore SPIN indicator进行了分析,这些指标体系不同程度地存在体系不完整或者操作性较弱的缺点,因而本文从指标的管理意义、数据的可获得性以及对于IT行业项目关键指标的影响程度三方面综合考虑,以IT行业项目管理三角形为依据建立了四类指标,十六个指标项为内容的IT行业项目关键指标体系,该指标体系包含了IT行业项目的主要管理目标以及影响这些目标的关键因素。论文对于指标依赖关系的研究则主要采用了定量分析的方法,对指标关系的研究又可以区分为线形关系和非线性关系研究。其中的线性研究采用简化指标关系的处理方式,给出了一种实际工作中容易操作,基于项目生产率、缺陷率和交付率三个派生指标的简化线性模型。而现实世界的IT行业项目质量和管理指标的依赖关系应该是非线性的依赖关系,因而非线性依赖关系的研究作为论文定量研究的主要内容。对于非线性关系的研究根据所研究的对象特点采用了不同的研究方法。因为对于IT行业项目质量状况研究往往缺乏客观数据,论文采用了可以综合处理主观经验数据和客观数据的贝叶斯信任网络模型。论文首先建立了IT行业项目质量的基本贝叶斯网络,通过参数模型预报IT项目中的“潜伏”缺陷数目,在此基础上,论文对项目质量的基本贝叶斯网络进行了扩展,提出了对于客户更为有意义的项目质量的扩展贝叶斯网络,该网络引入了IT项目评审、IT项目测试、IT系统的使用程度等重要网络节点,从而使得扩展后的贝叶斯网络可以为实际使用中的IT系统进行IT项目质量预报。为项目质量预报建立的贝叶斯网络还可以通过IT系统实际使用过程中所积累的缺陷信息为网络提供反馈,以有效修改网络中的先验概率,从而为后续的IT系统的质量预报提供更准确的参考信息,最终达到持续优化项目质量预报贝叶斯网络的目的。论文对于相对直观的IT项目管理指标的研究则采用了支持向量机算法模型。在对IT项目的工期指标和工作量指标的历史数据进行了归一化处理之后,将所研究的样本数据分割为训练数据、试验数据和检验数据。在进行支持向量机建模时主要应用了径向基核函数的向量机对于工期和工作量指标进行了回归模拟,并采用模拟模型对数据进行了预报。在使用支持向量机分析IT项目工期和IT项目工作量后,论文还对影响工期和工作量的相关指标进行了灵敏度分析,发现项目规模对于IT项目工作量的影响最为明显,项目团队人员规模位居其次,而项目类型、行业类型以及开发语言类型对于工作量的影响则相对轻微,甚至可以忽略不计；项目规模对于IT项目工期的影响程度则不如对工作量的影响更为明显,项目团队人员规模和项目类型对工期的影响仅次于项目规模,而行业类型和开发语言的影响则相对轻微。在论文中作者根据IT项目质量和管理指标的特点不同而选用了不同的研究方法,但其结论应该结合起来使用,以便为IT项目的量化管理提供决策参考依据。论文最后还介绍了作者在IT项目质量和管理指标体系应用方面所从事的咨询工作,指出IT项目质量和管理指标体系关系的建立和应用对于IT行业的可持续发展有着不可替代的重要作用。更多还原

【Abstract】 There usually exists typical difficulties for domestic IT project management work, such as time slippage, cost overrun, low-grade quality and so on. Based on careful examination, the thesis derives the conclusion that one of the root causes for the failed IT project is lack of quantitative method to establish objectives for the IT projects. Hence, the main goal of the thesis is to establish an IT project management indicator framework and explore the quantitative dependant relationship among the project quality and management indicators.The thesis combines the qualitative and quantitative method to study the IT project quality and management indicator framework, the qualitative method is mainly used to establish the management indicator framework. According to the results of careful examination of the literatures in the field of project management and author’s consulting practices in the field of IT quantitative project management, thesis explores three dominant IT project management models-PMBOK, ICB and CMMI. However, there is no definite guiding rule in these management models for establishing IT project management indicator framework. The thesis also analyses the industry indicator frameworks, such as ISO 15939, SEI indicator, ISBSG data questionnaire, SPR indicator and Bangalore SPIN indicator. Due to incompletency or lack of practice, these indicator frameworks are still not satisfying to describe the complete and practical management objectives for the IT project management. Therefore, the thesis designs a new framework for the IT project management indicator. Combining three factors, management concept of the indicator, data availability of the indicator and the important relation to the key indicator, the thesis suggests a 4-catogory 16-indicator framework. The suggested indicator framework is also based on project management triangle model, it encompasses 4 key indicators and other 12 affiliated indicators.The thesis uses quantitative method to study the dependency relationship among the indicators. The thesis assumes that there exist two types of dependant relationship, linear relationship and non-linear relationship. The linear model simplifies the indicator relationship and presents a simple and linear model in order to achieving better operation performance. However, the IT project management indicator relationship in the real world should be more complex and the relationship is non-linear. Therefore, the non-linear relationship among the indicator framework is the main content for the quantitative study method. Considering different features of the study objects, the thesis adopts two types of analysis model. Because there is usually lacking the IT system quality related data, the Bayesian Belief Network(BBN) model is selected for it can combine the pre-tested subjective information with post-tested objective information. The first step of modeling a BBN IT quality model is building a Basic BBN model, it can forecast the potential defect number. The second step of BBN modeling is to extend the basic BBN IT quality model to form the Extended BBN model, it introduces review node, testing node, usage frequency node to predict the potential defect number which will be found during the end user operation period. The BBN predicting model can also use the real defect number information and feedback it to the BBN model to adjust the corresponding probability of the pre-test condition. Therefore, the BBN model can optimize the model for predicating IT system defect number in a continuous manner.The thesis uses the Support Vector Machine (SVM) to study the IT project schedule indicator and effort indicator. The IT project schedule and effort related historical data are firstly normalized to meet the needs for SVM processing, then the data is divided up to 3 data files, including training sub-data, testing sub-data and predicting sub-data. Then the thesis adopts Radial Basis Function SVM to regress and predict the IT project schedule indicator and effort indicator。The thesis also analyses the sensitive degree among the indicators. Among the indicators which affect the IT project effort, the project size is the most important factor, the project team size is the second important. However, the contribution from project type, application type and the development language type are comparatively slight, even can be omitted. Project size contributes less to IT project schedule comparing with contribution to IT project effort, although it is still the most important one. The project team size and project type contribute similarly to the project schedule, the contributions from the application type and the language type are still very slight.Although the thesis applies different research methods, the research results should be linked together in order to provide information for IT project management decision work. Finally, the thesis also introduces author’s consulting practices in the field of IT quantitative project management, points out that IT project management indicator framework study and application is indispensable to continuous development for IT industry.更多还原