识别 GitHub 上的异常提交.pdf

1、本文档共35页，可阅读全部内容。
2、原创力文档（book118）网站文档一经付费（服务费），不意味着购买了该文档的版权，仅供个人/单位学习、研究之用，不得用于商业用途，未经授权，严禁复制、发行、汇编、翻译或者网络传播等，侵权必究。
3、本站所有内容均由合作方或网友上传，本站不对文档的完整性、权威性及其观点立场正确性做任何保证或承诺！文档内容仅供研究参考，付费前请自行鉴别。如您付费，意味着您自己接受本站规则且自行承担风险，本站不退款、不进行额外附加服务；查看《如何避免下载的几个坑》。如果您已付费下载过本站文档，您可以点击这里二次下载。
4、如文档侵犯商业秘密、侵犯著作权、侵犯人身权等，请点击“版权申诉”（推荐），也可以打举报电话：400-050-0827(电话支持时间：9:00-18:30)。

JOURNALOFSOFTWARE:EVOLUTIONANDPROCESS

J.Softw.Evol.andProc.0000;00:2–35

PublishedonlineinWileyInterScience().DOI:10.1002/smr

IdentifyingUnusualCommitsonGitHub

RamanGoyal,GabrielFerreira,ChristianKastnerandJamesHerbsleb

IndianInstituteofInformationTechnology,Allahabad

CarnegieMellonUniversity

SUMMARY

Transparentenvironmentsandsocial-codingplatformsasGitHubhelpdeveloperstostayabreastofchanges

duringthedevelopmentandmaintenancephaseofaproject.Especially,notiﬁcationfeedscanhelpdevelopers

tolearnaboutrelevantchangesinotherprojects.Unfortunately,transparentenvironmentscanquickly

overwhelmdeveloperswithtoomanynotiﬁcations,suchthattheyloosetheimportantonesinaseaof

noise.Complementingexistingprioritizationandﬁlteringstrategiesbasedonbinarycompatibilityandcode

ownership,wedevelopananomaly-detectionmechanismtoidentifyunusualcommitsinarepository,that

standoutwithrespecttootherchangesinthesamerepositoryorbythesamedeveloper.Amongothers,we

detectexceptionallylargecommits,commitsatunusualtimes,andcommitstouchingrarelychangedﬁle

typesgiventhecharacteristicsofaparticularrepositoryordeveloper.Weautomaticallyﬂagunusualcommits

onGitHubthroughabrowserplugin.Inaninteractivesurveywith173activeGitHubusers,ratingcommits

inaprojectoftheirinterest,wefoundthat,thoughourunusualscoreisonlyaweakpredictorofwhether

developerswanttobenotiﬁedaboutacommit,informationaboutunusualcharacteristicsofacommitchange

howdevelopersregardcommits.Ouranomaly-detectionmechanismisabu

数学毕业，从事大数据工作十几年，涉及二三十个行业

更多 >