Transcription of The Relationship Between Precision-Recall and ROC Curves
{{id}} {{{paragraph}}}
Example: marketing
The Relationship Between Precision-Recall and ROC Curves
TheRelationshipBetween Precision-Recalland ROC of ComputerSciencesandDepartment of BiostatisticsandMedicalInformatics,Unive rsity ofWisconsin-Madison,1210 WestDaytonStreet,Madison,WI,53706 USAA bstractReceiverOperatorCharacteristic(RO C)curvesarecommonlyusedtopresent ,whendealingwithhighlyskewed datasets, Precision-Recall (PR) Curves give a moreinformative pictureof an algorithm' show thata deepconnectionexistsbetweenROCspaceandPR space,such thata curve dominatesinROCspaceif andonlyif it corollaryis thenotionofanachievablePRcurve, which hasproper-tiesmuch like theconvex hullin ROCspace;we show ane cient , we alsonotedi erencesin thetwo types of Curves aresigni cant example,in PRspaceit is incorrectto ,algorithmsthatopti-mizetheareaundertheR OCcurve IntroductionIn machinelearning,current research hasshiftedawayfromsimplypresentingaccura cyresultswhenperform-inganempiricalvalid ationof al.
these spaces (Davis et al., 2005). The goal in ROC space is to be in the upper-left-hand corner, and when one looks at the ROC curves in Figure 1(a) they ap-pear to be fairly close to optimal. In PR space the goal is to be in the upper-right-hand corner, and the PR curves in Figure 1(b) show that there is still vast room for improvement.
Information
Domain:
Source:
Link to this page:
Please notify us if you found a problem with this document: