文档介绍：摘 要
软件项目是一个庞大的系统工程，相比以往的各种项目，由于其复杂性、时效性、技
术性、产品抽象性等等方面的原因，使得软件项目存在更多的不确定性。在这样的条件下，
大多数软件项目即使经过周密的计划，适当的控制也难以保证项目达到预期的目标，工期
延误、超出预算甚至完全失败等等都是项目管理者常常必须面对的现实。因此软件项目风
险在项目实践中越来越得到大家的重视，风险管理成为软件项目管理的重中之重，风险管
理的成功在很大程度上决定了软件项目的成功。而风险评估作为风险管理的重要组成部
分，对于风险管理有着至关重要的作用。
本文首先详细介绍了软件项目风险管理体系，综述了国内外软件风险管理的研究进
展，强调了风险评估对于风险管理的重要价值；然后，通过对Web化趋势下的软件项目的
风险分析，构建了适应未来发展需要的风险评估指标体系；接着将遗传算法与BP神经网络
相结合，解决了BP神经网络易陷入局部最优的问题，确定了本文采用的进化神经网络的方
法；并利用调查得到的样本数据，对该模型的有效性进行了验证，通过与当前常用的改进
BP神经网络的评估方法的对比，证实了该方法的优越性，从而为软件项目风险提供了一个
科学而有效的评估手段；最后针对当前某软件企业有意开发的软件项目进行风险的预测评
估，为项目决策提供依据。
关键词: 软件风险；风险评估；遗传算法；BP 算法；进化神经网络
Abstract
Software project can be considered as a huge systematic engineering. Compared to a variety
of former project, it has more uncertainties because of its complication, timeliness, technicality
and products abstraction. Even though most of software projects make careful plans and proper
controls, it is also quite difficult to ensure the desired objectives of the projects. Time delay and
an excess budget even a complete failure may be caused as a result. Therefore the risks of
software project earn widespread respect in project practice, and risk management goes to be the
key to software project management. To a great degree the success of risk management means
the success of software project. Assessment which is one part of risk management plays a crucial
role.
This research firstly introduces the risk management systems of software project, and
expounds the research progress in risk managent of software project both at home and abroad. It
also lays emphasis on risk assessment; moreover it constructs risk assessment index system
which may be adapted to both the present time and future; in addition on the basis of the above
index system, this paper integrates GA into BP neural network in order to cope with the problem
of falling into partial optimum, and the paper’s model of ENN is ascertain