ROC-Based Utility Function Maximization for Feature Selection and Classification with Applications to High-Dimensional Protease Data

In medical diagnosis, the diseased and nondiseased classes are usually unbalanced and one class may be more important than the other depending on the diagnosis purpose. Most standard classification methods, however, are designed to maximize the overall accuracy and cannot incorporate different costs...

Full description

Saved in:
Bibliographic Details
Published in:Biometrics 2008-12, Vol.64 (4), p.1155-1161
Main Authors: Liu, Zhenqiu, Tan, Ming
Format: Article
Language:English
Subjects:
Algorithms
Amino acids
Biometric Methodology
Biometrics
Biometry - methods
Biostatistics
Classification
Databases, Factual
Datasets
Diagnosis
Error rates
HIV 1
HIV Protease
Humans
Information classification
LASSO
Machine learning
Mathematical functions
Medical diagnosis
ROC
ROC Curve
Sensitivity
Software
Specificity
Statistical methods
Statistics, Nonparametric
Utility functions
Utility maximization
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:In medical diagnosis, the diseased and nondiseased classes are usually unbalanced and one class may be more important than the other depending on the diagnosis purpose. Most standard classification methods, however, are designed to maximize the overall accuracy and cannot incorporate different costs to different classes explicitly. In this article, we propose a novel nonparametric method to directly maximize the weighted specificity and sensitivity of the receiver operating characteristic curve. Combining advances in machine learning, optimization theory, and statistics, the proposed method has excellent generalization property and assigns different error costs to different classes explicitly. We present experiments that compare the proposed algorithms with support vector machines and regularized logistic regression using data from a study on HIV-1 protease as well as six public available datasets. Our main conclusion is that the performance of proposed algorithm is significantly better in most cases than the other classifiers tested. Software package in MATLAB is available upon request.
ISSN:0006-341X
1541-0420
DOI:10.1111/j.1541-0420.2008.01015.x