10.1080/17477778.2025.2536096">
 

Model validation levels: an automatable framework for model validation

Document Type

Article

Publication Date

10-6-2025

Abstract

This paper presents a detailed mathematical framework for computing a Model Validation Level (MVL), a metric for quantifying the amount of trust that can be placed in the results of a model. The framework is founded in three key pillars of validation: fidelity, referent authority, and scope. Each of these three pillars can be quantified to enable an objective and automatable comparison between model results and referent data. Combined, these pillars result in an MVL between zero and nine for the model that quickly communicates the model’s validity for a given intended use case. Additionally, the framework provides opportunities for identifying areas of model improvement to increase trust. The paper presents analysis on several metrics, demonstrating results for a variety of possible validation scenarios. Ultimately, the MVL framework provides an objective methodology for automating validation, enabling models to be quickly evaluated for trust in the ongoing digital transformation.

Comments

This article is published online ahead of inclusion in a future issue of Journal of Simulation


Source Publication

Journal of Simulation (ISSN 1747-7778 | eISSN 1747-7786)

This document is currently not available here.

Share

COinS