Important statistical considerations in classifier systems

Author

DeLeo, James M. ; Rosenfeld, Stephen J.

Author_Institution

Dept. of Clinical Res. Inf., Nat. Inst. of Health, Bethesda, MD, USA

fYear

2001

fDate

2001

Firstpage

285

Lastpage

293

Abstract

The performance of a classifier system may be limited due to the following: (1) nonmonotonic relationships between individual predictor co-factors and outcomes, (2) prevalence imbalances between development data and application environment data, and (3) failure to account for cost-gain economics. These issues are explored, and statistically-based techniques for treating them are presented. In addition, probabilistic and fuzzy interpretations of classifier outputs are discussed, a likelihood ratio transformation of classifier outputs is suggested and two new cost-gain indexes that rate classifier systems in global economic terms are introduced

Keywords

fuzzy logic; fuzzy set theory; inference mechanisms; pattern classification; performance index; statistics; uncertainty handling; application environment data; classifier outputs; classifier system performance; cost-gain economics; cost-gain indexes; development data; fuzzy interpretation; global economic rating; likelihood ratio transformation; nonmonotonic relationships; outcomes; predictor co-factors; prevalence imbalances; probabilistic interpretation; statistically-based techniques; Application software; Biomedical informatics; Economic forecasting; Electronic mail; Environmental economics; Extrapolation; Fuzzy systems; Hardware; Humans; Switches;

fLanguage

English

Publisher

ieee

Conference_Titel

Computer-Based Medical Systems, 2001. CBMS 2001. Proceedings. 14th IEEE Symposium on

Conference_Location

Bethesda, MD

ISSN

1063-7125

Print_ISBN

0-7695-1004-3

Type

conf

DOI

10.1109/CBMS.2001.941734

Filename

941734