# Recovering True Classifier Performance in Positive-Unlabeled Learning

arXiv: https://arxiv.org/abs/1702.00518 · HTML: https://mlsanity.com/paper/1702.00518/

## Key facts

- Cited by 5 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 2 Feb 2017.

## Built on by (most cited first, 5 of 5)

- [Uncertainty-aware Pseudo-label Selection for Positive-Unlabeled Learning](https://mlsanity.com/paper/2201.13192/) (2022)
- [Value Bonuses using Ensemble Errors for Exploration in Reinforcement Learning](https://mlsanity.com/paper/2602.12375/) (2026)
- [Improved proteasomal cleavage prediction with positive-unlabeled learning](https://mlsanity.com/paper/2209.07527/) (2022)
- [Learning from sanctioned government suppliers: A machine learning and network science approach to detecting fraud and corruption in Mexico](https://mlsanity.com/paper/2512.19491/) (2025)
- [Better Protein Function Prediction by Modeling Survivorship Bias](https://mlsanity.com/paper/2605.06879/) (2026)

## Abstract (arXiv)

A common approach in positive-unlabeled learning is to train a classification model between labeled and unlabeled data. This strategy is in fact known to give an optimal classifier under mild conditions; however, it results in biased empirical estimates of the classifier performance. In this work, we show that the typically used performance measures such as the receiver operating characteristic curve, or the precision-recall curve obtained on such data can be corrected with the knowledge of class priors; i.e., the proportions of the positive and negative examples in the unlabeled data. We extend the results to a noisy setting where some of the examples labeled positive are in fact negative and show that the correction also requires the knowledge of the proportion of noisy examples in the labeled positives. Using state-of-the-art algorithms to estimate the positive class prior and the proportion of noise, we experimentally evaluate two correction approaches and demonstrate their efficacy on real-life data.