缺陷数据的相似性度量方法改进  

Improvement of Similarity Measurement Method for Defect Data

作　　者：万琳[1] 杨腾翔 刘海宁[2] 

机构地区：[1]装甲兵工程学院信息工程系,北京100072 [2]中国兵器科学研究院,北京100089

出　　处：《计算机系统应用》2017年第8期152-156,共5页Computer Systems & Applications

摘　　要：模糊聚类分析主要研究样本的分类问题.本文利用模糊聚类方法对软件缺陷进行分类,引入缺陷数据属性权重计算方法,依据数据挖掘中的属性邻近性度量方法,对缺陷数据进行相似度分析.并按照属性类别进行分析,不仅体现了缺陷数据属性间的形贴近程度,而且体现了属性之间的距离贴近程度.本文方法对软件缺陷数据进行分析并对比度量结果,实验结果充分说明改进后的模糊聚类相似性度量方法在分类准确性方面有一定程度的提高.The study of fuzzy cluster analysis is mainly the classification of samples. In this paper, the fuzzy clustering method is used to classify the defects of software, and the method of attribute weight calculation is introduced. The similarity of defect data is analyzed with the method of attribute proximity in data mining. According to the category of attributes, it does not only reflect the degree of similarity between the attributes of the defect data, but also reflects the distance between the attributes. In this paper, the software defect data are analyzed and compared with the measurement results. The experimental results show that the improved fuzzy clustering similarity measurement method has somehow improved in classification accuracy.

关 键 词：模糊聚类 数据挖掘 软件缺陷 相似度 属性权重 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论]