Cost curve analysis of biometric system performance

Authors:
Mayra Sacanamboy;Bojan Cukic
Affiliations:
LANE Department of Computer Science and Electrical Engineering, West Virginia University, Morgantown, WV;LANE Department of Computer Science and Electrical Engineering, West Virginia University, Morgantown, WV
Venue:
BTAS'09 Proceedings of the 3rd IEEE international conference on Biometrics: Theory, applications and systems
Year:
2009

Citing 4
Cited 0

The relationship between Precision-Recall and ROC curves

ICML '06 Proceedings of the 23rd international conference on Machine learning
Cost curves: An improved method for visualizing classifier performance

Machine Learning
Validating a Biometric Authentication System: Sample Size Requirements

IEEE Transactions on Pattern Analysis and Machine Intelligence
Cost Curve Evaluation of Fault Prediction Models

ISSRE '08 Proceedings of the 2008 19th International Symposium on Software Reliability Engineering

Quantified Score

Hi-index	0.00

Visualization

Abstract

Biometric classification algorithms typically offer a range of performance characteristics which balance false non-match and false match rates. Nevertheless, the threshold which meets application requirements is usually selected without explicit consideration of cost implications of misclassification. This paper presents the analysis of recognition performance of multiple face and fingerprint algorithms using cost curves. Cost curves allow the introduction of misclassification costs and prior probabilities of proportions of genuine and impostor classes in the selection of biometric system thresholds. The inclusion of misclassification costs and prior probabilities is important since they can either change with time, or with the location where the biometric system is deployed.