面向单个文件的个性化缺陷预测方法  

Personalized Defect Prediction for Individual Source Files

作　　者：陈恒[1,2] 刘文广[1,2] 高东静[1,2] 彭鑫[1,2] 赵文耘[1,2] 

机构地区：[1]复旦大学软件学院,上海201203 [2]上海市数据科学重点实验室(复旦大学),上海201203

出　　处：《计算机科学》2017年第4期90-95,共6页Computer Science

基　　金：国家自然科学基金(61370079);国家高技术研究发展计划(863)(2013AA01A605)资助

摘　　要：现有的缺陷预测方法大多数是面向项目或个人的,这些方法或没有区分文件之间和开发人员之间的差异性,或只区分了开发人员的差异性。然而,在软件开发中,开发人员之间和代码文件之间的差异性是同时存在的,而且这些差异性都可能会影响缺陷建模或预测的结果。因此,如果缺陷预测方法忽视这些差异性或忽视其中任意一种,针对整个项目或某个开发人员建立缺陷预测模型均可能会影响预测准确性。针对此问题,提出了一种面向单个文件的个性化缺陷预测方法,即将每个开发人员修改每个代码文件的历史数据都作为单独的数据集,建立对应的缺陷模型,并将之用来预测对应开发人员修改对应文件的缺陷情况。通过实验初步确认了在单个文件的个人缺陷数据充分的情况下该方法能够有效地提高缺陷预测的准确性。Most defect prediction methods are project-oriented or developer-oriented.These methods do not distinguish the differences between source files and between developers,or only distinguish the differences between developers.However,there exist differences between developers and between source files in development process,and both differences may have an effect on defect prediction model and result.Therefore,if these two kinds of differences or either one kind of differences are ignored,and defect prediction models based on either the whole project or a single developer＇s development history are built,the prediction accuracy may be affected.To solve this problem,this paper proposed a personalized defect prediction method for individual source files,that is,we regarded the historical data of each file modified by each developer as an independent dataset,built a corresponding defect prediction model and used it to predict the defect possibility for the corresponding file modified by the corresponding developer.Experiments show the proposed method can improve the prediction accuracy with sufficient personal defect data.

关 键 词：缺陷预测 个性化 源代码 提交 

分 类 号：TP311.5[自动化与计算机技术—计算机软件与理论]