Adjusting the Outputs of a Classifier to New a Priori Probabilities: A Simple Procedure
Open accessThe DOI matches its Crossref record; 26 of the 26 references it cites were found in a registry or an index.
Abstract
It sometimes happens (for instance in case control studies) that a classifier is trained on a data set that does not reflect the true a priori probabilities of the target classes on real-world data. This may have a negative effect on the classification accuracy obtained on the real-world data set, especially when the classifier's decisions are based on the a posteriori probabilities of class membership. Indeed, in this case, the trained classifier provides estimates of the a posteriori probabilities that are not valid for this real-world data set (they rely on the a priori probabilities of the training set). Applying the classifier as is (without correcting its outputs with respect to these new conditions) on this new data set may thus be suboptimal. In this note, we present a simple iterative procedure for adjusting the outputs of the trained classifier with respect to these new a priori probabilities without having to refit the model, even when these probabilities are not known in advance. As a by-product, estimates of the new a priori probabilities are also obtained. This iterative algorithm is a straightforward instance of the expectation-maximization (EM) algorithm and is shown to maximize the likelihood of the new data. Thereafter, we discuss a statistical test that can be applied to decide if the a priori class probabilities have changed from the training set to the real-world data. The procedure is illustrated on different classification problems involving a multilayer neural network, and comparisons with a standard procedure for a priori probability estimation are provided. Our original method, based on the EM algorithm, is shown to be superior to the standard one for a priori probability estimation. Experimental results also indicate that the classifier with adjusted outputs always performs better than the original one in terms of classification accuracy, when the a priori probability conditions differ from the training set to the real-world data. The gain in classification accuracy can be significant.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
In the citing paper
“However, if wellcalibrated fraud probabilities are needed, then the posterior probabilities can be adjusted (Saerens et al., 2002).”
From Data engineering for fraud detection (2021) · cites this paper as Saerens et al., 2002
Citations
Cites
Maximum Likelihood from Incomplete Data Via the EM Algorithm
Dempster, A. P., Laird, N. M., & Rubin, D. B. (1977) · Journal of the Royal Statistical Society Series B: Statistical Methodology · doi:10.1111/j.2517-6161.1977.tb01600.x · Cited by 50,444
Probability, Random Variables, and Stochastic Processes.
Abrahams, J., & Papoulis, A. (1984) · Journal of the American Statistical Association · doi:10.2307/2288754 · Cited by 16,226
UCI Repository of machine learning databases
Blake, C. L. (1998) · Medical Entomology and Zoology · Cited by 10,582
The EM Algorithm and Extensions
Kushary, D. (1998) · Technometrics · doi:10.1080/00401706.1998.10485534 · Cited by 5,417
An Introduction to the Theory of Statistics
(1911) · Nature · doi:10.1038/087345a0 · Cited by 4,394
Probability, Random Variables, and Stochastic Processes
Papoulis, A., & Hoffman, J. G. (1967) · Physics Today · doi:10.1063/1.3034123 · Cited by 4,107
Journal of the Royal Statistical Society (B)
Fildes, R. (1993) · International Journal of Forecasting · doi:10.1016/0169-2070(93)90088-5 · Cited by 3,437
Text classification from labeled and unlabeled documents using EM
Nigam, K., Mccallum, A. K., Thrun, S., & Mitchell, T. (2000) · Machine Learning · doi:10.1023/a:1007692713085 · Cited by 2,736 · cited 1×
“For instance, Nigam et al. (2000) use the EM algorithm to learn classiers that take advantage of both labeled and unlabeled data.”
MethodDiscriminant Analysis and Statistical Pattern Recognition.
Hawkins, D. M., & McLachlan, G. J. (1993) · Journal of the American Statistical Association · doi:10.2307/2290355 · Cited by 2,633
Mixture models : inference and applications to clustering
McLachlan, G. J., & Basford, K. E. (1988) · M. Dekker eBooks · Cited by 2,094
Cited by
Items 301–362 of 362
Castro, L., Rueda, M. D. M., Sánchez-Cantalejo, C., Ferri, R., & Cabrera‐León, A. (2024) · Figshare · doi:10.6084/m9.figshare.26681119
Castro, L., Rueda, M. D. M., Sánchez-Cantalejo, C., Ferri, R., & Cabrera‐León, A. (2024) · Figshare · doi:10.6084/m9.figshare.26681122
Castro, L., Rueda, M. D. M., Sánchez-Cantalejo, C., Ferri, R., & Cabrera‐León, A. (2024) · Figshare · doi:10.6084/m9.figshare.26681116
Wu, X., 范, 瑞., Luo, T., & Hou, C. (2025) · Information Sciences · doi:10.1016/j.ins.2025.122706
Time Series Domain Adaptation: A Review
Furqon, M. T., Pratama, M., & Jatmiko, W. (2026) · Wiley Interdisciplinary Reviews Data Mining and Knowledge Discovery · doi:10.1002/widm.70130
The Resampling Delusion: A Geometric Theory of Class ImbalancePreprint
Sudjianto, A. (2026) · SSRN Electronic Journal · doi:10.2139/ssrn.6052514
Robust Label Shift Correction via Denoising Expectation-Maximization
Wu, X., Fan, R., & Hou, C. (2026) · Lecture notes in computer science · doi:10.1007/978-981-95-4987-0_5
Essais sur la modélisation du risque de crédit
Estran, R. (2020) · theses.fr (ABES) · doi:10.70675/1b91c4dfz2c50z4a82z9159z5c0f507c771a
Video compression of airplane cockpit screens content
Mitrica, I. (2021) · theses.fr (ABES) · doi:10.70675/18512bbdzfb2fz4993zaecfz8a0e94302a12 · Cited by 1
Model agnostic domain adaptation : application to fraud detection
Zhang, L. (2022) · theses.fr (ABES) · doi:10.70675/d8b95b8fz350bz4a5cz8313ze6b57fa9818e
A unified framework for label shift quantification
Dussap, B. (2024) · theses.fr (ABES) · doi:10.70675/8848fbe1ze611z43d3z937azf54fa35decf2 · Cited by 1
Zhang, K., Wang, B. L., Zhou, H., & Zhuang, X. (2026) · Medical Image Analysis · doi:10.1016/j.media.2026.104074
Unlabeled Target-Domain Calibration for Tabular Classifiers under Label Shift
Li, Y., Shi, J., & Cui, X. (2026) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp55912.2026.11464109
Ulyanin, M. Y. (2026) · SSRN Electronic Journal · doi:10.2139/ssrn.6625797
Learning Category-Aware Sensitivity for Federated Test-Time AdaptationPreprint
Zhang, Y. L., Deng, X., Yang, G., Viriri, S., Cao, J., & Yi, C. (2026) · Research Square · doi:10.21203/rs.3.rs-9453051/v1
Focused PU learning from imbalanced data
Zavitsanos, E., & Παλιούρας, Γ. (2026) · Data Mining and Knowledge Discovery · doi:10.1007/s10618-026-01264-1
“For this purpose, we used it to detect financial misstatements in annual reports, a task where PU data naturally arise.”
Continuous Sweep for Binary Quantification Learning
Kloos, K., Karch, J. D., Meertens, Q., Scholtus, S., & de Rooij, M. (2026) · Journal of Classification · doi:10.1007/s00357-026-09548-3 · Cited by 4
“Examples of this third group are the Saerens-Latinne-Decaestecker algorithm, i.e., SLD (Saerens et al., 2002), Distribution y-Similarity, i.e., DyS (Maletzke et al., 2019), and King & Lu’s algorithm (King & Lu, 2008).”
Zhang, J. W., LI, P., Zhang, G. Y., Gao, X., & Gong, P. (2026) · Computers & Electrical Engineering · doi:10.1016/j.compeleceng.2026.111222
Heart Failure Prediction & Risk Stratification using Machine LearningPreprint
Ali, S., Faqah, O., Leavitt, M. A., & Asghar, W. (2026) · Research Square · doi:10.21203/rs.3.rs-9498049/v1
Class-Conditioned Gaussian Mixture Modeling for Imbalanced Time Series Quantification
Kabir, M. S., Mithila, M. M. R., Ngu, A. H. H., Farias, M. C. Q., & Gao, B. J. (2026) · Lecture notes in computer science · doi:10.1007/978-981-92-1465-5_44
“…and its variants—Adjusted CC (ACC), Probabilistic CC (PCC), and Probabilistic ACC (PACC) treat quantification as a post-hoc correction of classifier outputs [4, 1], while EM-based procedures adjust classifier scores to new priors by iteratively re-estimating prevalences from unlabeled samples [17].”
Hopson, S., Ebbert, J. L., Urie, P. M., & Corte, D. D. (2026) · International Journal of Clinical Practice · doi:10.1155/ijcp/9940137 · Cited by 7
Graph-Theoretic Multi-View Feature Engineering for Interpretable BGP~Anomaly DetectionPreprint
Motaali, S., de Vergara, J. E. L., de Pedro, L., & González, I. (2026) · SSRN Electronic Journal · doi:10.2139/ssrn.7010298
Hossain, M. I., Sadat, S., Sifat, J. A., Alvi, M. W. A., & Rahman, M. A. (2026) · SSRN Electronic Journal · doi:10.2139/ssrn.7022843
Factorizable joint shift revisited
Tasche, D. (2026) · Foundations of Data Science · doi:10.3934/fods.2026019
“…without recourse to any label observations, in Section 3.3 we have presented a generalisation to label distribution estimation of the expectation maximisation (EM) approach for estimating target prior class probabilities under label shift (Saerens et al., 2002) and shown that it is fit for purpose.”
InfluentialMorales, V. H., Bru-Cordero, O., & Martínez-López, A. J. (2026) · International Journal of Environmental Research and Public Health · doi:10.3390/ijerph23070895
“Threshold tuning is particularly attractive in clinical settings [36,37].”
Detecting Performance Drift in AI Models for Medical Image Analysis Using CUSUM Chart
Prathapan, S., Samala, R. K., Hadjiyski, N., Petrick, N., Zamzmi, G., Nelson, B., Delfino, J. G., D’Haese, P.-F., Maldonado, F., Nguyen, P., et al. (2026) · Journal of Imaging Informatics in Medicine · doi:10.1007/s10278-026-02113-9
Geng, M., Chu, Y., Zhang, X., & Dong, J. (2026) · Lecture notes in computer science · doi:10.1007/978-981-92-2868-3_11
Climate-related hazards and issue attention to climate policy in the European Parliament
Buzogány, Á., Chiru, M., & Surdea‐Hernea, V. (2026) · West European Politics · doi:10.1080/01402382.2026.2702599
Zhang, Y., Gadhia, N. A., Karagiannis, G., & Smyrnakis, M. (2026) · ACM International Conference on Bioinformatics, Computational Biology, and Health Informatics (ACM-BCB) · doi:10.1145/3807503.3819471
“The procedure for estimating the test class proportion from the observed decision rate and the confusion matrix is a standard technique in prior shift adaptation [26].”
Ameli-Mazandarani, Z., Behnaz, M., Amirifard, H., Ebadifar, A., Mirmohammadsadeghi, H., Dalaie, K., & Kavousinejad, S. (2026) · Sleep And Breathing · doi:10.1007/s11325-026-03774-0
Consensus across limbic areas predicts motivational disengagementPreprint
Burk, D. C., & Averbeck, B. B. (2026) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.64898/2026.07.28.741314
Single Domain Generalization Benchmark: Current State and Challenges
Imbulgoda, K., Tennakoon, R., Chuah, W., Korevaar, S., Xia, F., & Bab‐Hadiashar, A. (2026) · International Journal of Computer Vision · doi:10.1007/s11263-026-02985-y
Pugantsov, A., Fabris, A., Susto, G. A., & Moreo, A. (2026) · Information Processing & Management · doi:10.1016/j.ipm.2026.105095
FedTTP: Test-Time Transformer Personalisation for Federated Learning as a Service
Aziz, G., Mistry, S., & Krishna, A. (2026) · IEEE International Conference on Web Services (ICWS) · doi:10.1109/icws72778.2026.00015
“Methods such as MEMO and T3A reduce entropy via augmentation or classifier adaptation [23], [24], while EM and BBSE estimate shifted label distributions to recalibrate predictions [25], [26].”
Ahmed, W., Rahaman, M. A., Roy, P., & Alam, P. (2026) · Frontiers in Artificial Intelligence · doi:10.3389/frai.2026.1928858 · Cited by 1
“Expectation–maximization prior re-estimation (Saerens et al., 2002) and black-box shift estimation (Lipton et al., 2018) are consistent under the assumption that the class-conditional distribution P ( x | y ) is invariant.”
Chao, L., & Shunsuke, M. (2026) · Progress in Disaster Science · doi:10.1016/j.pdisas.2026.100729
Avoiding common failures in AI for health and medicine
Salaudeen, O., Zhang, H., Kim, E., Singh, K., & Ghassemi, M. (2026) · Cell · doi:10.1016/j.cell.2026.07.025 · Cited by 1
Otabor, E., Hassan, A., Okunlola, A. O., Lam, J. R., Alomari, L., & Hamilton, M. (2026) · European Heart Journal - Digital Health · doi:10.1093/ehjdh/ztag147 · Cited by 2
“…of the label’s prevalence in the released training split to the predicted logit, which requires no outcome data from External validation and recalibration of EchoNext-Mini in MIMIC-IV the target setting; 23,24 and five-fold out-of-fold Platt scaling fitted separately for each label, which does.”
Cheng, C., You, N., Zhao, P., & Dong, J. (2026) · Remote Sensing of Environment · doi:10.1016/j.rse.2026.115694
Non-asymptotic analysis of label shift quantification: robustness and calibration
Lecestre, A. (2026) · Test · doi:10.1007/s11749-026-01046-3 · Cited by 1
“This work complements the contributions of Saerens et al. (2002), Alexandari et al. (2020), and Garg et al. (2020), offering a comprehensive perspective of MLLS in label shift estimation.”
Ponduru, P. S., Nandanavanam, P. P. V., & Ponduru, S. K. K. (2024) · International Journal of Scientific Research in Science and Technology · doi:10.32628/ijsrst25126427
Flow Exporter Provenance as a Major Confounder in Cross-Dataset IoT Intrusion Detection
Rassam, M. A., & Al-Asaly, M. S. (2026) · Mathematics · doi:10.3390/math14193549
PRISM: Prior-Robust Operating Points for Cross-Generator AI-Generated Image Detection
Pan, H.-H. (2026) · 2026 7th International Conference on Computer Vision and Data Mining (ICCVDM) · doi:10.1109/iccvdm71812.2026.11709258
Mou, L., Lu, Y., Wu, Z., Zhang, Q., & Pu, Z. (2026) · Xenotransplantation · doi:10.1111/xen.70171 · Cited by 1
“Class imbalance and class-dependent behavior of classifiers and high-dimensional hubness can bias nearest-neighbor-based inference [ 21–24 ].”
Li, S., Mo, D., & He, C. (2026) · Mathematics · doi:10.3390/math14193571
Prevalence Shift and ICU Mortality Miscalibration Across Institutions
Aman Chandra, H., Shivhare, S., & Sahu, S. (2026) · International Conference on Emerging Techniques in Computational Intelligence (ICETCI) · doi:10.1109/icetci71772.2026.11710759
CABLE-kNN: Cached and balanced local evidence for binary classification
Gul, N., Ali, A., Aldahmani, S., & Khan, Z. (2026) · Machine Learning with Applications · doi:10.1016/j.mlwa.2026.101034
Critical appraisal of artificial intelligence studies for prostate cancer detection on MRI
Pooch, E., Agrotis, G., Schoots, I. G., Emberton, M., Kasivisvanathan, V., & Hu, Y. (2026) · European Radiology · doi:10.1007/s00330-026-12915-8
Machine Learning Models Using Electronic Health Record Data to Predict Obstructive Sleep ApneaPreprint
Hwang, N. H., Mazzotti, D. R., Mishra, S., Westover, M. B., Hwang, D., THOMAS, R. J., Mignot, E. J.-M., Katwa, U., Shi, J. M., Yan, R., et al. (2026) · medRxiv · doi:10.64898/2026.10.05.26364806
Beyond Single-Beat Classification: Quantifying Arrhythmia in Long-Term ECG via Prevalence Estimation
de Medeiros, J. G. B., Luiz, L. R., Boas, G. S. V., Silva, R. D. C., Marcacini, R. M., & Silva, D. F. (2026) · Proceedings - IEEE Symposium on Computer-Based Medical Systems · doi:10.1109/cbms69103.2026.00025
“The EMQ algorithm [20] is an iterative approach that does not require a classifier based on a fixed threshold.”
Hasanov, E. E., Kadhi, A. E., Safarli, G., & Kadhi, N. E. Z. E. (2026) · Proceedings - IEEE Symposium on Computer-Based Medical Systems · doi:10.1109/cbms69103.2026.00209
“The framework operates on FreeSurfer/FastSurfer morphometry rather than end-to-end CNNs, and feature stability reduces but does not eliminate prevalence-induced degradation; future work will implement explicit recalibration [26].”
Hamuq Signal: multimodal Bitcoin range-outcome prediction using Llika blockchain embeddingsPreprint
Zuñiga, E. W. V., & Zhao, L. (2026) · SSRN Electronic Journal · doi:10.2139/ssrn.7619904
Fazakis, N., Kostopoulos, G., Kotsiantis, S. B., & Ragos, O. (2026) · Applied Sciences · doi:10.3390/app162010308
Dasarwar, P., & Hindayar, J. N. (2026) · Natural Hazards · doi:10.1007/s11069-026-08481-3
“…express predictions at the mapped-inventory prevalence, predicted probabilities p were adjusted with the prior-shift correction (King and Zeng 2001; Saerens et al. 2002 where π s is the proportion of landslides in the sample used to train the model and π t = 0.00196 is the mapped-inventory…”
InfluentialFederated Blockchain Learning to Support Root-Cause Analysis of Medical Incidents
Zouad, S., Khamer, L., Rached-Kanouni, M., Zaidi, S., Ayad, N., & Calafate, C. T. (2026) · Computers · doi:10.3390/computers15100736
Sengar, H. K., Trochin, Y., Dwivedi, T., Prasad, A., Kumar, A., & Goel, T. (2026) · Research Square · doi:10.21203/rs.3.rs-11236577/v1
Non-intrusive load monitoring of persistent single-phase loads in distribution networksPreprint
Aliyari, M., Hatami, M., Nizamani, A. G., & Ayele, Y. Z. (2026) · SSRN Electronic Journal · doi:10.2139/ssrn.7654999
Stabilizing EM-based quantification under label shift
González, P., Pérez-Mon, O., & del Coz, J. J. (2026) · Knowledge-Based Systems · doi:10.1016/j.knosys.2026.117259
Atef, S., & Eltawil, A. (2026) · Applied System Innovation · doi:10.3390/asi9110219
Adapting Skew-Probabilistic Neural Networks for Credit-Risk Probability Estimation
Bimoldan, M., & Kasenov, S. E. (2026) · Machine Learning and Knowledge Extraction · doi:10.3390/make8110352
Probability Calibration Under Class-Prevalence Shift: A Reproducible Multi-Dataset Stress Test
Ledwaba, K., & Mokoena, T. (2026) · Interdisciplinary Journal of AI Machine Learning & Data Science · doi:10.66261/73c9df20
Kluger, D. M. (2022) · Zenodo (CERN European Organization for Nuclear Research) · doi:10.5281/zenodo.6377525
Metadata
- Authors
- Saerens, M., Latinne, P., & Decaestecker, C.
- Affiliations
- Université Libre de Bruxelles (BE)
- Venue
- Neural Computation
- Publisher
- MIT Press
- Published
- 2002-01-01
- Volume
- 14
- Issue
- 1
- Pages
- 21-41
- ISSN
- 0899-7667 (ISSN-L)
- 1530-888X
- OpenAlex
- W2141253686
- Cited by
- 361
- References
- 26
- Citations by year
- 2026: 57 · 2025: 28 · 2024: 26 · 2023: 24 · 2022: 23 · 2021: 32 · 2020: 27 · 2019: 19 · 2018: 15 · 2017: 9 · 2016: 17 · 2015: 11 · 2014: 9 · 2013: 4 · 2012: 6
- FWCI
- 4.14
- Type
- journal article
- Language
- en
- Open access
- green
- dipot.ulb.ac.be
- Licence
- other-oa
- Locations
- Neural Computation (publishedVersion)
- PubMed (publishedVersion)
- Digital Access to Libraries (Université catholique de Louvain (UCL), l'Université de Namur (UNamur) and the Université Saint-Louis (USL-B)) (submittedVersion)
- Dépôt institutionnel de l'Université libre de Bruxelles (Université Libre de Bruxelles) (submittedVersion, other-oa)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Imbalanced Data Classification Techniques
- Physical Sciences › Computer Science › Artificial Intelligence › Machine Learning and Data Classification
- Physical Sciences › Computer Science › Artificial Intelligence › Data Stream Mining Techniques
- Keywords
- posterior probability, expectation-maximization algorithm, likelihood maximization, multilayer neural networks, case-control study, classification accuracy
- MeSH terms
- Classification, Probability, Probability Theory
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.