摘要
由于在线用户评论具有数据量大、质量良莠不齐等特点,单纯依靠人工校读方法发现其中的演化需求耗时耗力,也无法满足以用户为中心的软件版本快速更新的需要,因此提出一种基于用户评论的潜在演化需求发现方法 DICM.该方法首先对预处理后的用户评论文本进行基于信息增益的特征选择,接着使用下采样来降低训练集与测试集的不平衡度,最后利用朴素贝叶斯分类器分类出潜在演化需求,以辅助需求工程师进行演化需求的抽取.对照实验结果表明,使用DICM方法发现的潜在演化需求可以有效辅助需求分析师进行演化需求的获取,减轻需求分析师工作量并减小个体差异.同时,获得了关于DICM方法的用户可接受性及未来改进方向.
Since the online user comments have the features of large amount,varied quality of comment content etc,detecting the potential evolution requirements only by manual proofreading is labor-intensive and time-consuming,and also unable to meet the needs of user-centric software version quick update,thus a potential evolution requirements detect method DICM based on user comments was proposed.The method firstly selects the features from preprocessed user comments text based on information gain,then uses down-sampling to decrease the unbalancedness between training set and test set,finally applies the naive Bayes classifier to sort out the potential evolution requirements,to assist requirements analysts to extract evolution requirements.The results of control experiments carried out showed that the potential evolution requirements detected by DICM can assist the requirements analysts to elicit the evolution requirements effectively,decrease the manual effort and reduce the individual difference.Meanwhile,the user acceptance and some insights about the improvement of DICM also have been found.
出处
《武汉大学学报（理学版）》
CAS
CSCD
北大核心
2015年第4期347-355,共9页
Journal of Wuhan University:Natural Science Edition
基金
国家自然科学基金资助项目(61170026)
武汉市科技攻关计划项目(201212521826)
重庆市重点实验室专项基金项目(2012ECSC0210)
关键词
软件演化需求
用户评论
朴素贝叶斯分类器
下采样
software evolution requirements
user comments
naive Bayes classifier
down-sampling
