outlier detection
IRI: https://spec.industrialontologies.org/ontology/construct/OutlierDetection
Defined In: https://spec.industrialontologies.org/ontology/biopharma/Statistics/
SubClass Of: computing process
Class Hierarchy
owl:Thing › bfo:entity › bfo:occurrent › bfo:process › planned process › computing process › outlier detection
Definition
computing process in which input data are evaluated for the presence of outliers according to a detection criterion
Explanatory Notes
1) This class is intended to cover a broader range of outlier detection approaches than those currently represented in STATO, including both hypothesis-based statistical tests and non-hypothesis or algorithmic methods for identifying outliers. 2) Outlier detection is distinct from evaluation against a material specification, statistical control-chart rules, or an expected process trend. An observation may be flagged as a potential outlier without being out of specification, out of control, or out of trend, and those classifications do not by themselves establish that the observation is an outlier.
Examples
- applying a generalized extreme Studentized deviate test to replicate assay results to identify one or more potential outliers; applying an interquartile-range criterion to supplier-lot impurity results to flag atypical observations; applying robust multivariate-distance criteria to metabolite profiles to identify atypical culture samples; applying an isolation-based anomaly-detection algorithm to historical process data to flag observations for investigation
Adapted From
- https://www.sciencedirect.com/topics/mathematics/outlier-detection#:~:text=Outlier%20detection%20is%20defined%20as,fraud%20detection%20and%20data%20cleaning.&text=How%20useful%20is%20this%20definition?
Primitive Class
This class is declared primitive and it does not have necessary and sufficient condiftions defined.
Primitive Rationale
There are insufficient constructs to create a set of necessary and sufficient conditions.
Formal Axioms
First-Order Logic Axioms
DataTransformation(x) ∧ ∃y (achievesPlannedObjective(x, y) ∧ OutlierDetectionTestingObjective(y)) → OutlierDetection(x)
Semi-Formal Natural Language Axioms
if x is a ‘data transformation’ and x ‘achieves planned objective’ some ‘outlier detection testing objective’ then x is an ‘outlier detection’