简介：机器学习者的精确性被数据的质量影响学习者上被劝诱。在这篇论文，训练数据集的质量被移开划分过滤器作为吵闹检测的例子改进。合适的数据集首先被分成子集，；不同基础学习者在每这些裂口上被劝诱。如果它被某些基础学习者分类，预言以如此的一个方法被联合，一个例子作为吵闹被识别。划分过滤器的二个版本被使用：多重划分的过滤器；反复划分的过滤器。过滤器移开的例子的数字被过滤器的投票的计划调节；重复的数字。这研究的主要目的是比较造在上的最后模型的预兆的表演过滤；未过滤的训练数据集。一个高保证软件项目的软件测量数据的案例研究被执行。模型的预兆的表演在过滤合适的数据集上造了，这被显示出；在吵闹的测试数据集上评估了通常比造在上的那些好吵闹(未过滤)合适的数据集。然而，预兆的性能被噪音的存在在评估数据集基于某些好攻击的过滤器影响。

简介：Becausetrustisregardedasanessentialsecuredrelationshipwithinadistributednetworkenvironment,selectingservicesovertheInternetfromtheviewpointoftrusthasbeenamajortrend.CurrentresearchabouttrustmodelandevaluationinthecontextofWebservicesdoesnotrationallyandaccuratelyreflectsomeessentialcharacteristicsoftrustsuchassubjectiveuncertaintyanddynamism.Inthispaper,weanalyzesomeimportantcharacteristicsoftrust,andsomekeyfactorsthataffectthetrustrelationintheWebserviceenvironment.Accordingly,weproposeatrustmodelbasedonCloudModeltheorytodescribethesubjectiveuncertaintyoftrustfactors.Atime-relatedbackwardcloudgenerationalgorithmisgiventoexpressthedynamismoftrust.Furthermore,accordingtothetrustmodelandalgorithm,aformalizedcalculationapproachisprovidedtoevaluatethetrustdegreeofservicesrequestorsinproviders.Ourexperimentshowsthattheevaluationoftrustdegreecaneffectivelysupporttrust-decisionsandprovideahelpfulexploitationforselectingservicesbasedontheviewpointoftrust.

简介：Thequalityofasoftwaresystemispartiallydeterminedbyitsstructure(topologicalstructure),sotheneedtoquantitativelyanalyzethequalityofthestructurehasbecomeeminent.Inthispaperanovelmetriccalledsoftwarequalityofstructure(SQoS)ispresentedforquantitativelymeasuringthestructuralqualityofobject-oriented(OO)softwaresviabugpropagationanalysisonweightedsoftwarenetworks(WSNs).First,thesoftwaresystemsaremodeledasaWSN,weightedclassdependencynetwork(WCDN),inwhichclassesarenodesandtheinteractionbetweeneverypairofclassesifanyisadirectededgewithaweightindicatingtheprobabilitythatabuginoneclasswillpropagatetotheother.ThenweanalyzethebugpropagationprocessintheWCDNtogetherwiththebugpronenessofeachclass,andbasedonthis,ametric(SQoS)tomeasurethestructuralqualityofOOsoftwaresasawholeisdeveloped.TheapproachisevaluatedintwocasestudiesonopensourceJavaprogramsusingdifferentsoftwarestructures(oneemploysdesignpatternsandtheotherdoesnot)forthesameOOsoftware.Theresultsofthecasestudiesvalidatetheeffectivenessoftheproposedmetric.TheapproachisfullyautomatedbyatoolwritteninJava.