跳到正文
Patterns· Abu-Shehab·收录于· 6 天前

A Bayesian framework for anomaly detection in scientific data based on Benford’s law

A Bayesian framework for anomaly detection in scientific data based on Benford’s law

原文摘要

出版方登记 · OpenAlex

Ensuring data integrity is critical in modern data-intensive science, motivating the development of scalable methods for anomaly detection across datasets. Benford's law, which describes a characteristic distribution of leading digits in naturally occurring data, offers a domain-agnostic signal of irregularities, but its practical use is limited by difficulties in quantifying adherence and uncertainty, especially for datasets of varying sizes. We introduce simulation-based Benfordness estimation (SBBE), a Bayesian framework that models data as a mixture of Benford-conforming and non-conforming components. SBBE provides uncertainty-aware estimates, supports interpretability, and enables principled comparisons across datasets. We validate the approach using simulations and real-world Benford-compliant data and apply it to bioactivity datasets, where it revealed irregularities and sources of deviation. These results demonstrate the value of SBBE as a practical proxy for assessing data quality across scientific domains.

页面未能读取全文,以上为出版方登记的摘要;方法与图表见原文

来源:Patterns · doi.org