Cost-Sensitive Evaluation for Binary Classifiers
Pierangelo Lombardo, Antonio Casoli, Cristian Cingolani, Shola Oshodi, Michele Zanatta · 2025
Read Cost-Sensitive Evaluation for Binary Classifiers by Pierangelo Lombardo, Antonio Casoli, Cristian Cingolani, Shola Oshodi, Michele Zanatta (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2510-22016.txt.
Description
Pierangelo Lombardo, Antonio Casoli, Cristian Cingolani, Shola Oshodi, and Michele Zanatta argue that classifier scorecards should be built around what mistakes actually cost. Picking a metric shapes which model wins, how it is tuned, and whether it ships, yet no accepted measure directly targets total classification cost. Lopsided classes are often treated as a defect to fix even when fixing them raises real costs.
The team defines Weighted Accuracy, a version of accuracy where each example counts according to its cost. A general reweighting scheme that adapts any cost-style measure or loss to uneven classes without resampling the data. Because results are expressed in cost terms, scores from different datasets become comparable, and gaps between development data and the data a model will really face are handled openly.
Catalog record
| CREATOR | Pierangelo Lombardo, Antonio Casoli, Cristian Cingolani, Shola Oshodi, Michele Zanatta |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 49 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2510-22016