改进的软件模块化二元相似度测量（英文）

发布时间：2018-01-30 06:55

本文关键词： 二元相似度测量二元特征测量组合软件模块化　出处：《Frontiers of Information Technology & Electronic Engineering》2017年08期 　论文类型：期刊论文

【摘要】：目的:各种各样的二元相似度测量在聚类方法中被用来确定数据中的相似实体的同类组。这些相似度测量大多数仅基于特征的存在或缺失。二元相似度测量在软件模块化中亦能与不同的聚类方法一起用于提高软件系统的可理解性与可管理性。每种相似度测量都有其优势与不足,分别能使聚类结果优化或恶化。创新点:本文强调了软件模块化中一些已有的著名的二元相似度测量的优势。此外,基于这些已有的相似度测量,新提出了几种改进的相似度测量。方法:首先,介绍了一些软件模块化中已有的著名的二元相似度测量的优势。接着,提出了几种改进的新的相似度测量。结合具体例子,说明这些新方法整合了JCJNM和RR这几种已有的二元相似度测量的优势。最后,通过实验比较新方法与已有方法的结果,验证所提新方法的有效性。结论:实验结果表明相较于已有的相似度测量,本文所提出的新的二元相似度测量结果的可信度更高。这些新方法能减少任意决策的数量,增加聚类过程中聚类的数量。尽管这些新方法仅基于数据的二元特征向量表达,它们能被用来测试任何编程语言编写的软件系统。
[Abstract]:Objective:. A variety of binary similarity measurements are used in clustering methods to identify similar groups of similar entities in data. Most of these similarity measurements are based on the presence or absence of features. Binary similarity measurements are used in software modules. It can also be used to improve the understandability and manageability of software system together with different clustering methods. Each similarity measurement has its own advantages and disadvantages. Innovation: this paper emphasizes the advantages of some well-known binary similarity measurements in software modularization. In addition, based on these existing similarity measurements. Several improved similarity measurement methods are proposed. Firstly, the advantages of some well-known binary similarity measurements in software modularization are introduced. Several new improved similarity measurements are proposed. Combined with specific examples, these new methods integrate the advantages of JCJNM and RR binary similarity measurement. Finally. The effectiveness of the proposed method is verified by comparing the results of the new method with the existing methods. Conclusion: the experimental results show that compared with the existing similarity measurement. The new binary similarity measurement results proposed in this paper are more reliable. These new methods can reduce the number of arbitrary decisions. Although these new methods are based on binary eigenvector representation of data, they can be used to test software systems written in any programming language.
【作者单位】： Faculty
【基金】：supported by the Office of Research,Innovation,Commercialization and Consultancy(ORICC) Universiti Tun Hussein Onn Malaysia(UTHM),Malaysia(No.U063)
【分类号】：TP311.5
【正文快照】： 1 Introduction Clustering is an approach that makes clusters of similar entities in the data.Entities in a cluster are similar to each other(based on characteristics or features)while they are distinct from entities in other clusters.In the software dom

【相似文献】