Normalized excess cost (NEC) reveals that classifiers typically make fewer high-cost errors than raw error rates suggest, with cost-sensitive training providing benefits only when instance costs are predictable from input features.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Instance-Level Costs for Nuanced Classifier Evaluation
Normalized excess cost (NEC) reveals that classifiers typically make fewer high-cost errors than raw error rates suggest, with cost-sensitive training providing benefits only when instance costs are predictable from input features.