QQ客服 官方微博 反馈留言

题名	基于需求特征分析的新需求代码规模估算方法
作者	高健
学位类别	硕士
答辩日期	2010-05-27
授予单位	中国科学院研究生院
授予地点	北京
导师	王青
关键词	新需求 代码规模估算 需求特征分析 数据挖掘
学位专业	计算机软件与理论
中文摘要	需求演化频繁地发生在软件项目的生命周期中。如果能够对需求演化影响的代码规模进行准确的估算，就能够有效的帮助项目管理者分配项目资源，从而降低软件项目失败的风险。一般来说，需求演化分为增加新需求、修改需求和删除需求三种方式。对于需求的修改与删除所影响的代码规模，可以利用需求之间的依赖关系、需求与代码之间的关联关系以及代码之间的调用关系来进行估算。而对于增加的新需求的代码规模，则难以利用以上的关系进行估算。 因此本文提出两种基于需求特征分析的新需求代码规模估算方法，通过将需求进行需求特征提取，分析需求特征数据以估算新需求代码规模。其中一种为使用聚类的估算方法EMCNR-C，该方法将新需求与已有需求进行聚类，利用划分结果中处在同一簇的已有需求对新需求的代码规模进行计算。由于EMCNR-C方法的估算精度受聚类参数的影响较大，本文提出另一种使用加权调整的估算方法EMCNR-W，该方法根据需求之间的相似度计算全部的已有需求对新需求的影响权重，然后对新需求的代码规模进行计算。因为EMCNR-W方法避免了聚类参数带来的影响，所以在平均情况下能够对新需求的代码规模进行更为准确的估算。 基于以上两种估算方法，本文实现了一个自动化的新需求代码规模估算工具，该工具在输入需求特征数据后能够对新需求代码规模做出相对准确的估算。本文使用实例对两种估算方法进行对比分析，结果表明EMCNR-C在聚类参数选择适当的情况下能够对新需求代码规模做出较为准确的估算；而EMCNR-W方法则在平均情况下能够对新需求代码规模做出较为准确的估算。
英文摘要	Requirements evolve frequently during the software development lifecycle. If the impacted code size can be accurately estimated, it will help the project managers effectively allocate the resources and reduce the risk of project failure. There are three forms of requirement evolution in general: new requirement addition, requirement changes and requirement removed. To estimate the impact of requirement changes and deleted, the relationships between the existing code and requirements should be used, which include the dependent relationship between the requirements, the traceable relationship between the requirements and the code, and the function called relationship in the code. However, it is hardly to estimate the impact of the new requirements by using the relationships above. In order to resolve the problem, this thesis presents two estimation methods for the code size of new requirements based on requirement feature analysis. After extracting requirements feature, this method can estimate the code size of new requirements by analyze the requirement feature data. One of the methods uses clustering (EMCNR-C). This method uses a clustering algorithm to partition the new requirements and the history requirements, and uses the similar history requirements to estimate the code size of the new requirements. Considering the parameter of the clustering algorithm impacts the estimation effect strongly. Another method uses weight adjustment (EMCNR-W). This method calculates the impact weights between the history requirements and the new requirements by using the requirement similarity. It can use all the history requirements to estimate the code of the new requirements. This method avoids the problem of selecting a clustering parameter, so it can get a more accurate estimation. Based on the two estimation methods above, this thesis presents an automatically analysis tool for estimate the code size of new requirements. This thesis does a controlled experiment between the EMCNR-C and EMCNR-W methods by using a real project. The results show that the EMCNR-C can do accurately estimation when the clustering parameter is appropriate. In addition, the EMCNR-W method does better than the EMCNR-C on average.
语种	中文
学科主题	软件工程
公开日期	2010-06-07
内容类型	学位论文
源URL	[http://124.16.136.157/handle/311060/2327]
专题	软件研究所_互联网软件技术实验室 _学位论文
推荐引用方式 GB/T 7714	高健. 基于需求特征分析的新需求代码规模估算方法[D]. 北京. 中国科学院研究生院. 2010.

个性服务

查看访问统计

相关权益政策

暂无数据

收藏/分享

除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。