Effect of features burdened with classical error, misclassifications, Berkson error, and group-summary assignments on machine learning performance and interpretability
Abstract
Measurement error is widespread in modern data pipelines, but its influence on predictive modeling and variable importance assessments in machine learning is often underestimated. Our study examines how input measurement error affects both prediction accuracy and interpretability across four common additive error types: classical error; misclassification of predictive features; Berkson error, caused by algorithmic or model-based assignment errors; and group-summary (aggregation) assignment errors. Using controlled Monte Carlo simulation experiments that cover regression and classification, we evaluate five popular machine learning models: generalized linear models, support vector machines, random forests, XGBoost, and multilayer perceptrons, across various error levels. To distinguish the effect of the error from optimization issues, hyperparameters are tuned separately for each dataset, and performance is measured using standard predictive metrics and permutation feature importance. Results show that measurement error generally reduces predictive performance and can significantly alter the ranking and estimated contribution of highly predictive features, with degradation patterns that vary by error type and model. These findings clarify when performance declines might be mistaken for distributional shifts, among other things, and highlight the importance of diagnosing and correcting measurement error before drawing conclusions about model robustness, generalization, or variable relevance.
Related
- Lobar analysis of classical and deep learning radiomic features in chronic airflow limitation
- Quantum-classical hybrid learning framework for Parkinson’s disease prediction using acoustic and clinical features
- Positron Emission Tomography-Computed Tomography (PET-CT) Diagnosis of Large-Vessel Giant Cell Arteritis Presenting as Pyrexia of Unknown Origin Without Classical Cranial Features: A Case Report.
- Predicting Robot-Based Tracking Performance After Stroke Using Lesion Features and Clinical Assessments: A Pilot Study
- Functional performance tests in male and female professional dancers from four classical ballet companies – a cohort study
BibTeX
@article{mohamed2026effect,
title = {{Effect of features burdened with classical error, misclassifications, Berkson error, and group-summary assignments on machine learning performance and interpretability}},
author = {Zeinab Mohamed and Tamer Oraby},
journal = {Scientific Reports},
year = {2026},
month = oct,
doi = {10.1038/s41598-026-74435-1},
url = {https://www.nature.com/articles/s41598-026-74435-1},
note = {CC BY-NC-ND 4.0},
}
RIS
TY - JOUR
TI - Effect of features burdened with classical error, misclassifications, Berkson error, and group-summary assignments on machine learning performance and interpretability
AU - Zeinab Mohamed
AU - Tamer Oraby
PY - 2026
DA - 2026/10/11
JO - Scientific Reports
DO - 10.1038/s41598-026-74435-1
UR - https://www.nature.com/articles/s41598-026-74435-1
AB - Measurement error is widespread in modern data pipelines, but its influence on predictive modeling and variable importance assessments in machine learning is often underestimated. Our study examines how input measurement error affects both prediction accuracy and interpretability across four common additive error types: classical error; misclassification of predictive features; Berkson error, caused by algorithmic or model-based assignment errors; and group-summary (aggregation) assignment errors. Using controlled Monte Carlo simulation experiments that cover regression and classification, we evaluate five popular machine learning models: generalized linear models, support vector machines, random forests, XGBoost, and multilayer perceptrons, across various error levels. To distinguish the effect of the error from optimization issues, hyperparameters are tuned separately for each dataset, and performance is measured using standard predictive metrics and permutation feature importance. Results show that measurement error generally reduces predictive performance and can significantly alter the ranking and estimated contribution of highly predictive features, with degradation patterns that vary by error type and model. These findings clarify when performance declines might be mistaken for distributional shifts, among other things, and highlight the importance of diagnosing and correcting measurement error before drawing conclusions about model robustness, generalization, or variable relevance.
SN - 2045-2322
ER -