Highlights:<\/p>\n\n\n\n
- \n
- Traditional scoring rules may favor misclassifications over correct predictions.<\/li>\n\n\n\n
- Proposes \u201cSuperiority\u201d to ensure correct predictions score better than incorrect ones.<\/li>\n\n\n\n
- PBS and PLL modify traditional scoring rules to penalize misclassifications more.<\/li>\n\n\n\n
- Proves PBS and PLL remain strictly proper while satisfying the Superior property.<\/li>\n<\/ul>\n\n\n\n
Abstract:<\/p>\n\n\n\n
This study introduces novel superior scoring rules called Penalized Brier Score (PBS<\/em>) and Penalized Logarithmic Loss (PLL<\/em>) to improve model evaluation for probabilistic classification. Traditional scoring rules like Brier Score and Logarithmic Loss sometimes assign better scores to misclassifications in comparison with correct classifications. This discrepancy from the actual preference for rewarding correct classifications can lead to suboptimal model selection. By integrating penalties for misclassifications, PBS<\/em> and PLL<\/em> modify traditional proper scoring rules to consistently assign better scores to correct predictions. Formal proofs demonstrate that PBS<\/em> and PLL<\/em> satisfy strictly proper scoring rule properties while also preferentially rewarding accurate classifications. Experiments showcase the benefits of using PBS<\/em> and PLL<\/em> for model selection, model checkpointing, and early stopping. PBS<\/em> exhibits a higher negative correlation with the F1 score compared to the Brier Score during training. Thus, PBS<\/em> more effectively identifies optimal checkpoints and early stopping points, leading to improved F1 scores. Comparative analysis verifies models selected by PBS<\/em> and PLL<\/em> achieve superior F1 scores. Therefore, PBS<\/em> and PLL<\/em> address the gap between uncertainty quantification and accuracy maximization by encapsulating both proper scoring principles and explicit preference for true classifications. The proposed metrics can enhance model evaluation and selection for reliable probabilistic classification.<\/p>\n\n\n\n
Graphical abstract<\/h2>\n\n\n\n
![\"\"](\"https:\/\/norc.aut.ac.ir\/wp-content\/uploads\/2025\/03\/1-s2.0-S0888613X25000623-gr001_lrg-1024x409.jpg\")
<\/figure>\n\n\n\nRef: https:\/\/doi.org\/10.1016\/j.ijar.2025.109421<\/a><\/p>\n\n\n\n
The full version of the paper \u201cSuperior scoring rules for probabilistic evaluation of single-label multi-class classification tasks\u201d by Rouhollah Ahmadian, Mehdi Ghatee, & Johan Wahlstr\u00f6m is now available online in \u201c<\/a>International Journal of Approximate Reasoning<\/a>\u201d. It can be downloaded by clicking the following Share Link before May 14, 2025. No sign up, registration, or fees are required.<\/p>\n\n\n\n