QQ客服 官方微博 反馈留言

题名	基于需求变更的软件缺陷预测方法
作者	何磊
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	李明树
关键词	缺陷预测，需求度量，信息检索，支持向量机
学位专业	其他专业
中文摘要	软件缺陷对于软件质量以及软件项目的成本等都有非常重要的影响。软件缺陷预测技术从20世纪70年代发展至今,一直是软件工程领域最活跃的内容之一,在分析软件质量、控制软件成本方面起着重要的作用。软件缺陷预测技术分为静态和动态两种，其中基于软件度量元的静态缺陷预测技术是现今比较成熟且应用比较广泛的技术，它通过对缺陷相关的软件产品，例如代码，进行度量元的提取和计算，建立预测的模型以预测后期可能引入的缺陷。随着数据挖掘技术的成熟，越来越多的静态缺陷预测方法开始对软件项目历史数据进行分析和建模，挖掘这些历史数据以预测缺陷被证明是更加精确和可靠的。 但是，现有缺陷预测技术大都需要对软件的设计、代码或者测试等相关的活动进行分析，无法在软件生命周期的早期活动，例如需求活动，通过预测这些活动引起的潜在的缺陷的分布、类型和规模，从而为软件过程的后续活动提供早期预警以及有意义的依据和参考。本文提出了一种基于需求变更的软件缺陷预测方法，这种方法以迭代开发的升级性项目为应用对象，使用信息检索和数据挖掘相关技术，分析和处理升级项目中的历史需求文档和缺陷记录数据，建立支持向量机（SVM）分类预测模型，从而对后续版本的需求变更可能引入的缺陷进行预测。 本文深入细致的研究了现有的缺陷预测技术，分析并对比了这些技术的应用范围、特点和局限性，在此基础上提出了一种新的基于需求变更的软件缺陷预测方法。本方法使用信息检索技术关联匹配软件项目历史需求和历史缺陷，并根据历史需求所关联的缺陷分布属性将这些需求分类，之后对需求进行特征即度量元的提取和计算，从而建立SVM分类模型。当新的需求变更发生时可以使用建立的模型预测其分类，以此预测可能引入的缺陷。本文还介绍了基于需求变更的缺陷预测系统的核心功能设计与实现，并在最后通过使用一个实际的商业软件项目数据集对方法和系统进行了验证。在实验中预测系统表现了较高的精确度，可以提供较为可靠的缺陷预测结果，这些预测结果可以为软件项目中需求开发提供有效的变更影响分析，为控制软件成本和项目风险提供有效的决策支持。
英文摘要	Software defect is one of the most important items we are concern about during the process of software engineering. Defects have really close connection with software quality and cost. Since 1970’s, technologies of predicting defect have been developed rapidly for analyzing software quality and controlling software cost. Defect prediction technologies can be divided into static and dynamic ones. Static defect prediction approaches, which are widely used nowadays, are based on different phases of software process using defect-related data metrics to establish prediction models. With the development of data mining technology, more and more researches are to discover information and patterns from historical software data repositories. Predicting defects by mining history data is proved to be more precisely and reliable. However, existing defect prediction methods rely heavily on software code metrics available during late software life cycle phase to establish the estimation relationship with software defects. This thesis proposes a novel approach to predict defects based on the change of requirement by adopting Information Retrieval technique and Support Vector Machines (SVM). This study aims to propose a method to predict defects of a new version in a software product series using information available early from requirement phase. The method takes advantage of a continuous and disciplined software measurement framework across the previous software releases, and leverages on the collected historical requirement and defect data. This thesis studies the current methods of Defect Prediction Technologies. Base on the analysis of these technologies, this research provides and implements a novel approach to predict defects based on requirement change. More specifically, the method extracts requirements metrics from historical requirement documents, adopts SVM to establish requirement classification model based on various requirement and metrics, and predict defects introduced by new upgrade requirements based on the classification models and their corresponding defect profiles. Experiment results on actual data of a series of upgrade projects show that by analyzing the historical requirements and defects data, the prediction accuracy can reach a significant level. The proposed method may provide useful insight for understanding the impact of an upgrade requirement, as well as good facilitation for controlling project risks and assuring software trustworthiness.
语种	中文
学科主题	软件工程
公开日期	2010-06-03
内容类型	学位论文
源URL	[http://124.16.136.157/handle/311060/2292]
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	何磊. 基于需求变更的软件缺陷预测方法[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。