【作者】 郭琦；

【导师】 马培军；

【作者基本信息】 哈尔滨工业大学 ， 计算机技术， 2014， 硕士

【摘要】 随着软件产业的发展，需求工程在软件开发中的作用逐渐显现出来，如何准确地获取雇主的需求，是保证软件项目成功实施的重要前提。软件系统规模越来越大，想要充分获取到用户的需求，必然会导致需求数据量增大的问题；另一方面，软件系统会涉及多个雇主，每个雇主都期待软件系统有完善的功能、快捷的效率，而这些雇主对需求的选取可能持有不同的意见，如何选择一个能反映不同雇主需要的需求集合，成为了一个亟待解决的问题。目前，国内对于需求优选的研究尚处于初级阶段，没有一个比较完善的系统来解决多雇主的需求优选问题。本文将解决多雇主的需求优选中需求量庞大、需求重复、需求选取的问题，利用基于搜索的需求优选技术，从需求集合中提取出近似最优的需求集合，以确保在现有资源下，尽可能满足多个雇主的要求。本文介绍了多雇主的需求优选系统的设计和实现过程。该系统包括需求去冗余；雇主聚类；需求优选和方案排序四个功能模块。需求去冗余模块去除初始需求中的重复信息；雇主聚类将关注相同需求的雇主进行聚类，从而有效减少多目标优化问题中的优化目标数量，更有效的解决实际需求优选中的多雇主问题；需求优选模块将雇主类对需求的选择作为一个优化目标，采用基于搜索的多目标优化算法，搜索在有限资源下，使所有雇主都达到满意的需求集合；方案排序模块对需求优选模块得到的需求选择方案进行排序，以便决策制定者更好的进行选择。本文在原有的基于搜索的需求优选技术上进行了改进，将雇主类替代雇主，作为需求优选的目标，采用基于文档的NSGA-II算法解决需求优选问题，有效降低了多目标优化中目标的数量，降低了计算复杂度，此外，还增加了需求去冗余的处理，采用自然语言处理的相关技术，对初始需求进行预处理，减少了需求的数量。实验结果表明，本文实现的需求优选系统，达到了预计的目标，改进的基于搜索的需求优选技术更加有效。更多还原

【Abstract】 With the development of the software industry, requirements engineering plays amore and more important role in software engineering. How to obtain requitementsfrom a large amount of stakeholders accutately is the prerequisite to implement asoftware project successfully. With the imcreasing of the scale of software, if you wantto meet the user’s needs, will inevitably lead to increase the amount of requirements, onthe other hand, a software system may involve many stakeholders. Every stakeholderhopes the system to be perfect and efficiency, who may take different ideas. How tochoose requirements which can fulfill all stakeholders’ needs is a problem.At present, the research of requirements selection still in its infancy, there is not aperfect system to solve multi-stakeholder requirements selection so far. The paper willsolve the multi-stakeholder requirements selection with a large amount of requirementswhich contain redundant information. We introduce a search-based technique forautomated analysis the problem of requiremtents selection for multiple stakeholders,with the limited resources.The paper describes the design and the implementation of the multi-stakeholderrequirements selection system. There are four functions, as follow: removing redundantrequirements, clustering for stakeholders, requirements selection and sort for solution.The system will delete redundant requirements from the initial requirements collectedfrom multiple stakeholders. The system will classify the stakeholders according to therequirements they focused on automatically. Then we can reduce the amount ofobjective effectively. So that, we can solve the problem more effectively. Inrequirements selection phrase, each cluster of stakeholders is considered as an objectivein their own right, with the help of search-based technique, to find the approximateoptimal set of requirements which fulfill all stakeholders’ satisfaction. At last, sort forsolution, determine a sort for solution to help decision makers make better choices forrequirements set.In this paper, we improved the original search-based technique formulti-stakeholder requirements selection. Each cluster of stakeholders is considered as aobjective rather than a stakeholder, it will reduce the number of targets to be optimized,so it can reduce the computational complexity. Then we use archive-based NSGA-II tosearch optimal solutions. Besides, the paper removed redundant requirements to reducethe amount of requirements with natural language processing technology. Theexperimental results show that, the multi-stakeholder requirements selection systemreached the expected target, the improved search-based technology proposed in this paper is more effective.更多还原