Ten Simple Rules to build a Model Life Cycle

This is a Preprint and has not been peer reviewed. This is version 1 of this Preprint.

Add a Comment

You must log in to post a comment.


Comments

There are no comments or no comments have been made public for this article.

Downloads

Download Preprint

Authors

Timothée Poisot , Daniel J Becker, Cole B Brookson, Ellie Graeden, Sadie Jane Ryan, Gemma Turon, Colin J Carlson

Abstract

Just like data, models have their own life cycle. By recognizing how one’s model fits within the life cycle of the data (or at least, ensuring that the model life cycle is understood), we can identify opportunities to foster new collaborations, encourage better practices in data analysis, and ultimately accelerate research. In this manuscript, we introduce the Model Life Cycle and develop a series of ten simple rules aimed at facilitating collaborations between data collectors, curators, users, and modellers, as well as maximizing the potential for re-use of models. We explore the idea of a Model Life Cycle, starting from the assumption that it will address machine learning (ML) models, i.e. models that can be trained and deployed iteratively, and whose focus is on prediction of quantifiable phenomena. Specifically, we are interested in clarifying the use of models in large, interdisciplinary groups, where the actual modelling exercise may involve only a subset of the group (e.g., with others collecting and standardizing data).

DOI

https://doi.org/10.32942/X2XG8V

Subjects

Bioinformatics, Ecology and Evolutionary Biology, Software Engineering

Keywords

model life cycle, data life cycle, Iterative forecasting

Dates

Published: 2024-02-10 03:58

Last Updated: 2024-02-10 08:58

License

CC BY Attribution 4.0 International

Additional Metadata

Language:
English

Data and Code Availability Statement:
Not applicable