软件开发与维护中需求变更应对策略与bug管理的研究与实现

发布时间：2018-08-31 20:49

【摘要】：本文通过分析传统的需求和bug管理系统中的数据记录,完成对需求的相似性判断推荐和需求变更影响的衡量以及bug修复优先级、严重性、报告重复性等预测推荐,来提高传统系统中历史数据的使用率和用户的体验,并将新增的预测推荐功能和传统系统结合的基础上进行了设计与实现。本文的主要研究点包括以下四点:(1)传统文本分类算法针对短文本分类的改进,传统的文本分类算法是基于大量冗余文本的基础而设计的,而短文本所含冗余信息有限,若仍然使用传统的文本分类方法一般达不到预期结果。本文短文本的固有特点进行分析,从提高文本内联性和关键冗余扩展两方面着手,进行改进实验,并证实改进后的算法能有效提高短文本分类的准确率。(2)对需求变更影响进行衡量。定义系统实现组件和需求间的关系、组件重要性、需求变更影响度以及相似需求的判定推荐算法来实现需求变更影响的衡量,并在系统中提供用户评价反馈功能来评估系统算法,以系统评估代替人工的方式,达到统一了评估标准和降低了人工工作量的目的。(3)对bug修复的优先级,严重性和bug报告的重复性进行预测判断。通过数据挖掘推荐算法的使用和改进,对开源系统的bug记录进行实验判断。通过系统进行bug的修复预测,解决了新产生bug的属性的判定用户经验和主观性影响大的问题,并且可以进行批量预测,有效提高bug归属判断的预测效率。(4)在传统的需求与bug管理系统的基础上增加了推荐预测功能,并基于用户个人标签和用户系统标签来进行任务归属的推荐。在分类预测算法中主要借鉴近年来的一些研究成果,并在其基础上做了相应的适用性改进以解决短文本的文本向量矩阵数据稀疏性问题,并有效了提高推荐的准确率,系统对用户标签的自动提取,可以进行个性化推荐功能。
[Abstract]:Based on the analysis of traditional requirements and data records in bug management system, this paper accomplishes the evaluation of requirement similarity and requirement change, as well as bug repair priority, severity, report repeatability, and so on. To improve the utilization rate of historical data and user's experience in the traditional system, and design and implement the new predictive recommendation function combined with the traditional system. The main research points of this paper are as follows: (1) the traditional text classification algorithm is based on the foundation of a large number of redundant text, and the redundant information contained in the short text is limited. If the traditional text classification method is still used, the expected results are generally not achieved. This paper analyzes the inherent characteristics of the short text, from the aspects of improving the inline of the text and extending the key redundancy, the improvement experiment is carried out. It is proved that the improved algorithm can effectively improve the accuracy of short text classification. (2) the impact of requirement change is measured. The relationship between components and requirements, the importance of components, the degree of impact of requirement change and the recommendation algorithm of similar requirements are defined to measure the impact of requirement change. And provide the function of user evaluation feedback in the system to evaluate the system algorithm, system evaluation instead of manual way to achieve the purpose of unifying the evaluation standard and reducing the manual workload. (3) the priority of bug repair. Severity and repeatability of bug reports were predicted. Through the use and improvement of data mining recommendation algorithm, the bug records of open source system are judged experimentally. Through the repair prediction of bug, the problem of judging the attributes of bug is solved, which has great influence on user experience and subjectivity, and the batch prediction can be carried out. The prediction efficiency of bug attribution judgment is improved effectively. (4) based on the traditional requirement and bug management system, the recommendation prediction function is added, and the task attribution recommendation is made based on the user personal label and the user system label. In the classification prediction algorithm, we mainly draw lessons from some research achievements in recent years, and on the basis of it, we make the corresponding applicability improvement to solve the sparse problem of text vector matrix data of short text, and effectively improve the accuracy of recommendation. The system can automatically extract user tags, and can carry out personalized recommendation function.
【学位授予单位】：北京邮电大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP311.5

【参考文献】