On the Path to an Ideal ROC Curve: Considering Cost Asymmetry

Francis Bach

Description

On the Path to an Ideal ROC Curve: Considering Cost Asymmetry in Learning Classifiers Francis Bach David Heckerman Eric Horvitz UC Berkeley Microsoft Research January 2005

• correct classification


• linear classification


• false positive


• roc curves


• classification using


• training linear


• asymmetries


• considering cost


• asymmetric testing

Sujets

• Microsoft
• Horvitz
• Linear classifier
• Donna Kossy

Informations

• Publié par: pefav
• Nombre de lectures: 57
• Langue: English

Extrait

On the Path to an Ideal ROC Curve: Considering Cost Asymmetry

in

Francis Bach UC Berkeley

Learning Classifiers

David Heckerman Eric Horvitz Microsoft Research

January 2005

•

•

•

•

Outline

Asymmetric testing cost and ROC analysis

Training linear classifiers

Efficient algorithm to vary the training cost asymmetry

Mismatch between training and testing asymmetries