文献类型：期刊文献

中文题名：一种特征转移和域自适应的异质缺陷预测方法

英文题名：Feature Transfer and Domain Adaptation Approach for Heterogeneous Defect Prediction

作者：黄燕[1];徐贤[1];虞慧群[1];杨星光[1]

机构：[1]华东理工大学信息科学与工程系,上海200237

年份：2022

卷号：43

期号：1

起止页码：186

中文期刊名：小型微型计算机系统

外文期刊名：Journal of Chinese Computer Systems

收录：CSTPCD;;北大核心:【北大核心2020】；CSCD:【CSCD_E2021_2022】；

基金：国家自然科学基金项目(61872142,62072299,61772336,61702334,61772200,61173048,61557218,61572318)资助;上海市信息化发展专项基金项目(201602008)资助;上海市浦江人才计划项目(17PJ1401900)资助;上海市自然科学基金项目(17ZR1406900,17ZR1429700)资助;华东理工大学教育研究基金项目(ZH1726108)资助;上海市2019年度科技创新行动计划项目(19511101403)资助。

语种：中文

中文关键词：异质缺陷预测;特征迁移;域自适应;类不平衡

外文关键词：heterogeneous defect prediction;feature transfer;domain adaption;class imbalance

摘要：软件缺陷预测是软件工程领域中的一个研究热点.跨项目缺陷预测(Cross-Project Defect Prediction, CPDP)采用源项目的缺陷数据来预测目标项目的缺陷倾向性.然而,源项目和目标项目的特征可能会有所不同.因此,研究人员提出了异质缺陷预测(Heterogeneous Defect Prediction, HDP).为了提高HDP模型的性能,本文提出了一种基于特征迁移和域自适应的异质缺陷预测(Feature Transfer and Domain Adaptation, FTDA)方法.首先,FTDA使用基于相关性的特征选择算法从源项目的特征集中选择最优的特征子集.随后,使用欧氏距离进行匹配特征,并为每个目标项目选择最合适的源项目.再次,使用TCA算法解决不同项目之间的分布差异问题.最后,使用SMOTETomek算法处理类不平衡问题.为了验证FTDA方法的有效性,本文对AEEEM,PROMISE,NASA和Relink数据集中的24个项目进行了实证研究.实证研究表明,FTDA显著提高了异质缺陷预测的性能.
Software defect prediction is an active topic in the field of software engineering.Cross-project defect prediction(CPDP)adopts the defect data set of the source project to predict the defects of the target project.However, the metrics of the source and target projects may be different.Therefore, researchers proposed Heterogeneous Defect Prediction(HDP)method.To improve the performance of the HDP model, we propose a new HDP method called Feature Transfer and Domain Adaptation(FTDA).Firstly, FTDA uses the CFS(Correlationbased Feature Selection)algorithm to select an optimal subset from the metric set of the source project;Then, the Euclidean Distance is used not only to match features but also to select a suitable source project;Thirdly, the TCA algorithm is used to make the data distributions of the source and the target projects similar.Finally, the SMOTETomek algorithm is used to deal with class imbalance.To verify the effectiveness of the FTDA method, we conducted empirical study on 24 projects in the AEEEM,PROMISE,NASA,and Relink datasets.Empirical research shows that FTDA indeed outperforms the previous work.Therefore, we recommend using the FTDA method to build an HDP model.