Combining Data Preprocessing Methods With Imputation Techniques for Software Defect Prediction

Q4 Computer Science

International Journal of Open Source Software and Processes Pub Date : 2018-01-01 DOI:10.4018/IJOSSP.2018010101

Misha Kakkar, Eleni Constantinou, Apostolos Ampatzoglou, G. Robles, Jesus M. Gonzalez-Barahona, Daniel Izquierdo-Cortazar

{"title":"Combining Data Preprocessing Methods With Imputation Techniques for Software Defect Prediction","authors":"Misha Kakkar, Eleni Constantinou, Apostolos Ampatzoglou, G. Robles, Jesus M. Gonzalez-Barahona, Daniel Izquierdo-Cortazar","doi":"10.4018/IJOSSP.2018010101","DOIUrl":null,"url":null,"abstract":"SoftwareDefectPrediction(SDP)modelsareusedtopredict,whethersoftwareiscleanorbuggy usingthehistoricaldatacollectedfromvarioussoftwarerepositories.Thedatacollectedfromsuch repositories may contain some missing values. In order to estimate missing values, imputation techniquesareused,whichutilizesthecompleteobservedvaluesinthedataset.Theobjectiveof thisstudyis to identify thebest-suitedimputationtechniqueforhandlingmissingvalues inSDP dataset.Inadditiontoidentifyingtheimputationtechnique,theauthorshaveinvestigatedforthemost appropriatecombinationofimputationtechniqueanddatapreprocessingmethodforbuildingSDP model.Inthisstudy,fourcombinationsofimputationtechniqueanddatapreprocessingmethodsare examinedusingtheimprovedNASAdatasets.Thesecombinationsareusedalongwithfivedifferent machine-learningalgorithmstodevelopmodels.Theperformanceof theseSDPmodelsarethen comparedusingtraditionalperformanceindicators.Experimentresultsshowthatamongdifferent imputationtechniques,linearregressiongivesthemostaccurateimputedvalue.Thecombination oflinearregressionwithcorrelationbasedfeatureselectoroutperformsallothercombinations.To validatethesignificanceofdatapreprocessingmethodswithimputationthefindingsareappliedto opensourceprojects.Itwasconcludedthattheresultisinconsistencywiththeaboveconclusion. KeyWORDS Feature Selection, Instance Selection, Missing Value Imputation, Software Defect Prediction","PeriodicalId":53605,"journal":{"name":"International Journal of Open Source Software and Processes","volume":"4 1","pages":"1-19"},"PeriodicalIF":0.0000,"publicationDate":"2018-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"7","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"International Journal of Open Source Software and Processes","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.4018/IJOSSP.2018010101","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q4","JCRName":"Computer Science","Score":null,"Total":0}

引用次数: 7

Abstract

SoftwareDefectPrediction(SDP)modelsareusedtopredict,whethersoftwareiscleanorbuggy usingthehistoricaldatacollectedfromvarioussoftwarerepositories.Thedatacollectedfromsuch repositories may contain some missing values. In order to estimate missing values, imputation techniquesareused,whichutilizesthecompleteobservedvaluesinthedataset.Theobjectiveof thisstudyis to identify thebest-suitedimputationtechniqueforhandlingmissingvalues inSDP dataset.Inadditiontoidentifyingtheimputationtechnique,theauthorshaveinvestigatedforthemost appropriatecombinationofimputationtechniqueanddatapreprocessingmethodforbuildingSDP model.Inthisstudy,fourcombinationsofimputationtechniqueanddatapreprocessingmethodsare examinedusingtheimprovedNASAdatasets.Thesecombinationsareusedalongwithfivedifferent machine-learningalgorithmstodevelopmodels.Theperformanceof theseSDPmodelsarethen comparedusingtraditionalperformanceindicators.Experimentresultsshowthatamongdifferent imputationtechniques,linearregressiongivesthemostaccurateimputedvalue.Thecombination oflinearregressionwithcorrelationbasedfeatureselectoroutperformsallothercombinations.To validatethesignificanceofdatapreprocessingmethodswithimputationthefindingsareappliedto opensourceprojects.Itwasconcludedthattheresultisinconsistencywiththeaboveconclusion. KeyWORDS Feature Selection, Instance Selection, Missing Value Imputation, Software Defect Prediction

查看原文本刊更多论文

软件缺陷预测的数据预处理与Imputation相结合

SoftwareDefectPrediction(SDP)modelsareusedtopredict、whethersoftwareiscleanorbuggy usingthehistoricaldatacollectedfromvarioussoftwarerepositories。Thedatacollectedfromsuch repositories_可能_包含_一些_缺失的_值。> > order > >估计> >缺失的值，> > impuation> > techniquesareused，whichutilizesthecompleteobservedvaluesinthedataset。Theobjectiveof thisstudyis to identify_ thebest-suitedimputationtechniqueforhandlingmissingvalues inSDP数据集。Inadditiontoidentifyingtheimputationtechnique，theauthorshaveinvestigatedforthemost appropriatecombinationofimputationtechniqueanddatapreprocessingmethodforbuildingSDP模型。Inthisstudy，fourcombinationsofimputationtechniqueanddatapreprocessingmethodsare examinedusingtheimprovedNASAdatasets。Thesecombinationsareusedalongwithfivedifferent machine-learningalgorithmstodevelopmodels。Theperformanceof theseSDPmodelsarethen comparedusingtraditionalperformanceindicators。Experimentresultsshowthatamongdifferent imputationtechniques，linearregressiongivesthemostaccurateimputedvalue。Thecombination oflinearregressionwithcorrelationbasedfeatureselectoroutperformsallothercombinations。To validatethesignificanceofdatapreprocessingmethodswithimputationthefindingsareappliedto opensourceprojects.Itwasconcludedthattheresultisinconsistencywiththeaboveconclusion。关键词特征选择，实例选择，缺失值插值，软件缺陷预测

本文章由计算机程序翻译，如有差异，请以英文原文为准。

求助全文

约1分钟内获得全文求助全文

来源期刊

International Journal of Open Source Software and Processes Computer Science-Software

CiteScore

1.90

自引率

0.00%

发文量

期刊介绍： The International Journal of Open Source Software and Processes (IJOSSP) publishes high-quality peer-reviewed and original research articles on the large field of open source software and processes. This wide area entails many intriguing question and facets, including the special development process performed by a large number of geographically dispersed programmers, community issues like coordination and communication, motivations of the participants, and also economic and legal issues. Beyond this topic, open source software is an example of a highly distributed innovation process led by the users. Therefore, many aspects have relevance beyond the realm of software and its development. In this tradition, IJOSSP also publishes papers on these topics. IJOSSP is a multi-disciplinary outlet, and welcomes submissions from all relevant fields of research and applying a multitude of research approaches.