摘要
一个适合的软件项目风险管理系统对于软件开发的成功有着至关重要的作用。传统的软件项目风险管理系统往往采用人工的方法，已经不能满足软件开发的需要。因此，建立一个高效的软件项目风险管理系统有着重大的意义。
在调查和分析了风险管理、软件项目风险管理系统、建模技术的研究现状之后我们发现：在风险管理领域，信息系统正取代传统的手工方式逐渐成为软件项目风险管理系统的主流，在风险管理的理论基础上，当今的建模领域为风险管理建模提供了很多经典实践。但是这些建模分析技术对如何整合其他软件项目管理系统缺乏相应的实例，对于软件项目风险管理系统在异构系统环境下的应用缺乏解决方案。而系统工程领域和项目开发过程领域的风险管理建模方法都比较抽象化和一般化，对于建立一个符合软件开发特点的高效的软件项目风险管理系统还缺乏系统、具体的解决方案。
为了给软件项目建立一个能够有效支持软件项目风险管理、有效综合与协调异构系统环境，我们分析了软件开发过程的特征，将贝叶斯网络应用到软件项目风险管理系统中，并且针对贝叶斯网络在软件项目风险管理系统中应用的几个关键问题确定了相应的解决方案。在此基础上，我们综合风险管理、软件项目风险管理系统、建模技术等领域的过程方法，提出了一个适合于软件项目风险管理的、以贝叶斯网络、敏捷设计原则为架构基础的软件项目风险管理系统，建立了关键实践体系，并基于这一体系描述了框架，进行了具体软件项目风险管理系统的定义。
在理论研究的基础上，我们将此架构应用到一个软件项目风险管理系统的开发活动中，取得了良好的研究效果，该系统捕获风险数据的效率比应用其他建模技术的风险管理系统所提供的效率要高，系统与系统之间的整合度也比较高，研究后的调查反映基于贝叶斯网络的风险管理系统具备良好的效率和可操作性，敏捷设计原则也在实践中得到了验证和改进。
A proper software project risk management system is significant to successful software development. Traditional software project risk management system usually adopts manual methods, and cannot satisfy the needs of software development. Therefore, it’s of great significance to establish an efficient software project risk management system.
By investigating and analyzing the latest research of Risk Management, Software Project Risk Management System and Modeling Technologies, we find that Information System is becoming a mainstream of software project risk management system instead of manual methods in the field of Risk Management. Based on risk management, the field of Modeling Technologies provides many successful practices for modeling risk management. But these modeling and analyzing technologies and methods provides neither instance of integrating other software project management systems, nor solution of applying software project risk management system under multi-system environment. Likewise, the process methods in the fields of Systems Engineering and Software Development are too abstract and generalized, lacking solution of establishing an efficient software project risk management system.
In order to establish a software project risk management system for software development that can effectively support software project risk management and effectively integrate and coordinate multi-system activity, we analyzed the characteristic of software development process, applied Bayesian Networks Method to software project risk management system and developed solutions for key problems while doing so. Furthermore, we proposed a system for software project risk management, by integrating Bayesian Networks Methods, and Agile-based architecture, defining the key practices system, architecture and concrete processes of the software project risk management system.
Based on the theoretical research, we applied architecture to the development activities of a software project risk management system, and achieved good practical effect. The efficiency of the project was much higher than the average of the systems which are applied by other modeling technologies. The integrations between systems were kept in a very high level. The survey after pilot-running showed that the BN-based software project risk management system has good efficiency and applicability. Agile Design Methods also was verified and improved in practice.
引文
【1】 Roger S. Pressman. 软件工程实践者的研究方法. 机械工业出版社，2005
【2】 曹光忠. 软件项目的风险管理. 计算机时代，2005 (06)
【3】 斯蒂夫.迈克康奈尔. 快速软件开发----有效控制与完成进度计划. 电子工业出版社，2002
【4】 Cooper G F. Computational complexity of probability inferenceusing Bayesian Belief Networks[J]. Artificial Intelligence, 1993, 15;246-255.
【5】 Shenoy P P, Shafer G.. Axioms for probability and belief functionpropagation[J].In Uncertainty in artificial intelligence IV.Shachter R D,et al.eds. North-Holland. Amsterdam.1990, 169-198.
【6】 Bayes T R.Transactions. An Essay Towards Solving a Problem in theDoctrine of Chances[J].Philosophical 1763, 53; 370-418.
【7】 Adnan Darwiche. Conditioning algorithms for networks[C]. InProceedings of the 11th Conference and approximate inference in causalon Uncertainty in Artificial Intelligence (UAI), 1995, 99-107.
【8】 Pear J. Fusin, Propagation and structuring in beliefnetwork[J].Artificial Intelligence,1986, 29;241-288
【9】 Lauritzen S L, Spiegelhalter D J. and their Applications to ExpertLocal Computations with Probabilities on Graphical StructuresSystems[J]. Journal of the Royal Statistical Society, 1988
【10】 Kjaerulff U. Nested junction trees. [R] Uncertainty in ArtificialIntelligence: Proceedings of the Thirteenth Conference, (1997);294-301.
【11】 Shachter R D, D'Ambrosio B D, Del Favero B D, Symbolicprobabilistic inference in belief networks[C]. In Proc.Bm NationalConference on Artificial Intelligence, 1990, 126-131.
【12】 Dechter R. Bucket elimination: A unifying framework forreasoning[D]. Department of computer and Information ScienceUniversity of California,1999.
【13】 Li Zhaoyu, D .Ambrosio B. Efficient Inference in Bayes NetworksAs A Combination Problem[J]. International Journal of ApproximateReasoning, 1994.
【14】 Shachter,R. Probabilistic inference and influence diagrams[J].Operations Research, 1988 36;589-604.
【15】 Cooper G F. Computational complexity of probabilisticnetworks(Research Note)[J].Artificial Intelligence, 1990,inferenceusing Bayesian belief 42(2);393-405.
【16】 Henrion M. Propagating Uncertainty in Bayesian Networks byProbabilistic Logic Sampling[J], In Uncertainty in ArtificialIntelligence 2,1988,149-164.
【17】 Kjaerulff U. Nested junction trees. [R] Uncertainty in ArtificialIntelligence: Proceedings of the Thirteenth Conference, 1997;294-301.
【18】 David Poole. Average-case analysis of a search algorithm forestimating prior and posterior probabilities in Bayesian networkswith extreme probabilities[C]. In Proc.13th International Joint Conf.On Artificial Intelligence, 1993.
【19】 Poole D. The Use of conflicts in searching Bayesian networks[R],Proceedings of the Ninth Conference on Uncertainty in AI, WashingtonD.C, 1993,359-367.
【20】 Santos E J, Shimony S E. Deterministic nets[J].IEEE Transactionson Systems, Man, and Cybernetics,of marginal probabilities in bayes1998,28(4):377-393.
【21】 Kalev Kask,Javier Larrosa and Mina Dechter.Up and DownMini-Buckers: A Scheme for Approximating Combinatorial OptimizationTasks.
【22】 Friedman N, Geiger D, Goldszmidt M. Bayesian networkclassifiers[J]. Machine Learning,1997,29;131-161.
【23】 LamW, BacchusF. Learning Bayesian belief networks: an approachbased on the MDL principle [J].ComputationalIntelligence,1994,10;269-293.
【24】 Cheng die, Be11D, LiuWei-ru. Learning Bayesian networks from data;an efficient approach based on information-theory [J]. ArtificialIntelligence,2002,137(1,2);43-90.
【25】 Pear13. Probabilistic reasoning inintelligent systems: networksof plausible inference [M]. San Mateo CA: Morgan Kaufmann, 1988,117-133.
【26】 Meilir Page-Jones. UML 面向对象设计基础. 人民邮电出版社，2001
【27】 Robert C. Martin. 敏捷软件开发----原则、模式与实践. 清华大学出版社，2003
【28】 Erich Gamma, Richard Helm, Ralph Johnson, John Vlissides. 设计模式----可复用面向对象软件的基础. 机械工业出版社，2000
【29】 Joseph Schmuller. UML基础、案例与应用. 人民邮电出版社，2002
【30】 Granville Miller. UML2.0新特性. 非程序员，2004
【31】 Lauritzen S L, Spiegelhalter D 1. and their Applications to Expert LocalComputations with Probabilities on Graphical Structures Systems[J]. Journal ofthe Royal Statistical Society, 1988
【32】 Acid, S. and Campos, L.M., An algorithm for finding minimum d-Separatingsets in belief networks[R], Proceedings of the twelfth Conference of Uncertaintyin Artificial Intelligence, 1996.
【33】 Eric J. Naiburg, Robert A. Maksimchuk. UML数据库设计应用. 人民邮电出版社，2002
