基于改进的TF-IDF软件测试错误信息分析方法  

Method of software testing error analysis based on improved TF-IDF

作　　者：王茹[1,2] 严明[1] 王柳舒 

机构地区：[1]西安建筑科技大学信息与控制工程学院,西安710055 [2]西安建筑科技大学土木工程学院,西安710055

出　　处：《计算机应用》2016年第A02期259-261,共3页journal of Computer Applications

基　　金：国家自然科学基金资助项目(51278400)

摘　　要：针对软件测试领域人工分析测试用例错误信息工作量大、时间效率低的问题,提出了一种基于改进的词频-逆文本词频(TF-IDF)软件测试错误信息文本分析方法。首先,根据错误信息文本的特点对目标错误信息文本进行预处理,减少了干扰信息,缩短了计算时间;然后,结合关键词集合、TF-IDF和向量空间模型(VSM)计算文本特征向量,其中关键词集合避免了多次对数据库中错误信息文本进行TF-IDF权值计算,提高了计算效率;接着,利用余弦相似计算目标错误信息文本与数据库文本之间的相似度,并对相似度排序,从而找到相似度最高的错误信息,进而找到相关联的变更请求(CR);最后,自动关联CR。实验结果表明,该方法在软件测试错误信息分析方面能够有效提高时间效率。In order to solve these problems such as the heavy workload and low time efficiency of analyzing test case error text by hands, a new method of software testing error analysis based on the Terra Frequency-Inverse Document Frequency （ TF- IDF） was proposed. Firstly, in order to reduce the interference information and short the run time, the feature of error text was used to preprocess the target error text. Secondly, the feature vector of the error text was calculated by the key words collection, TF-IDF and Vector Space Model （VSM）. The TF-IDF weight calculation times of error texts in database were avoided by key words collection to improve calculation efficiency. Thirdly, the text similarity between database text and target error text was calculated by the cosine similarity, then sorted. The error text with greatest similarity in database and the related CR （ Change Request） were found. Lastly, the related CR was associated. Experimental results show that the proposed method can effectively improve the time efficiency in software testing error information analysis.

关 键 词：向量空间模型 TF-IDF 文本相似度量 余弦相似 软件测试 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]