The Çebi Method Anomaly detection and correction in a univariate number series — a parameter-free, scale-invariant, two-sided method
Year: 2026
Type: preprint
Abstract: A closed-form, parameter-free method for detecting and correcting anomalies — extreme values that break away from the body of a univariate number series. Both tails are pulled inward toward a band derived entirely from the series’ own structure; the method needs no threshold, training, or labels and is scale-invariant, order-invariant, and deterministic. Keywords anomaly detection · outlier correction · robust statistics · Winsorization · expected shortfall · tail estimation · parameter-free · scale-invariant
Author: ÇEBİ, AVNİ
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7166540986
Cites:
Cited by:
Related to: