Controlled experiments show class separability moderates oversampling effectiveness more strongly than imbalance ratio, challenging the IR-threshold paradigm.
Should we really use post-hoc tests based on mean-ranks?
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
-
Beyond Imbalance Ratio: Data Characteristics as Critical Moderators of Oversampling Method Selection
Controlled experiments show class separability moderates oversampling effectiveness more strongly than imbalance ratio, challenging the IR-threshold paradigm.