Intelligent decision support system for COTS based software development

基于商用现货（COTS）的软件开发（CBSD）过程依赖于识别、评估和选择满足系统需求和满足涉众偏好的COTS产品。关注第二个目标的方法主要基于共识方法，在这种方法中，涉众必须首先就解决方案评估标准及其相对重要性达成一致。然后，开发团队必须根据解决方案评估标准评估不同候选COTS的功能，以选择最合适的COTS。然而，这种方法存在几个问题：(1)在许多情况下，不可能协调利益相关者的关注点并就解决方案评估标准达成一致；(2)可能缺乏尝试不同情景来处理利益相关者偏好冲突的手段；(3)一个（或多个）利益相关者在谈判过程中改变其偏好的情况通常没有得到处理；(4)忽视利益相关者之间的联盟形成；(5)评价过程中固有的不确定性和相互依赖性有时得不到解决；(6)以前评估过程中用于重用信息的存储库要么集中，要么被忽略；(7)利益相关者重要性不同的情况往往被忽略。在这项研究中，我们提出了一种完全不同的方法，其中：(a)不需要初始共识；(b)上述缺陷已得到充分解决；(c)经验存储库是分散的，几乎不需要维护。多智能体系统（MAS）和基于本体的知识管理（OKM）的使用是本项目的独特之处。使用MAS实现系统提供了灵活性、可扩展性和对用户隐藏复杂性，并允许智能消息传递和知识级别协商。OKM允许知识的动态分类和分布，适应多样化的内容、表示和个性化的风格；结合有效的检索机制；并且几乎不需要集中管理和维护。

Commercial Off-The-Shelf (COTS) Based Software Development (CBSD) process relies on identifying, evaluating and selecting COTS products that satisfy system requirements, and satisfy stakeholders' preferences. The methodologies that focus on the second target are mostly based on the consensus approach in which the stakeholders have to first agree on the solution evaluation criteria and their relative importance. Then, the developer team has to evaluate the functionality of different COTS candidates against the solution evaluation criteria to select the most suitable COTS. However, there are several problems with such approaches: (1) in many cases it not possible to harmonize concerns of the stakeholders and agree on the solution evaluation criteria; (2) means for trying out different scenarios to deal with conflicting stakeholders' preferences may be missing; (3) situations where one (or more) stakeholder changes his/her preferences during the negotiation process is not usually dealt with; (4) coalition formation among stakeholders is ignored; (5) uncertainty and interdependencies inherent in the evaluation process are sometimes left unaddressed; (6) repositories for reuse of information from previous evaluation processes are either centralized or left out; and (7) cases where stakeholders have differing degree of significance are usually ignored. In this research we suggest a radically different approach in which (a) initial consensus is not required; (b) the above deficiencies are fully addressed; and (c) experience repositories are decentralized and require little or no maintenance. The use of multi-agent systems (MAS) and ontology-based knowledge management (OKM) are unique points in this project. Implementing the system using MAS offers flexibility, expandability and hiding complexity from users and allows for intelligent message passing and knowledge level negotiation. OKM allows for dynamic classification and distribution of knowledge, adapting to diversified contents, representations and personalized styles; incorporates efficient retrieval mechanisms; and requires virtually less or no centralized management and maintenance.