Sensitivity of machine learning regression models to data structure and quality in crop yield prediction
Mahoukpégo Luc Zinzinhédo, Mèvognon Firmin Mitchozounnou, Kolawolé Valère Salako, Romain Glèlè Kakaï · 2026
Read Sensitivity of machine learning regression models to data structure and quality in crop yield prediction by Mahoukpégo Luc Zinzinhédo, Mèvognon Firmin Mitchozounnou, Kolawolé Valère Salako, Romain Glèlè Kakaï (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/plos-pone-0353938.txt.
Description
Mahoukpégo Luc Zinzinhédo and co-authors stress-test crop yield prediction models by feeding them thousands of deliberately flawed datasets. Real farm data arrive with gaps and tangled measures, and advice on cleaning them is scattered. The simulation varied dataset size, predictor count, missing values, and overlap among predictors, crossed with cleaning choices, across seven learning methods, then repeated the checks on five real crop datasets.
Missing data hurt accuracy badly, tangled predictors faked one kind of score while worsening another, and bigger datasets reliably helped. Filling gaps by regression worked best overall, and one forest method stayed steadiest across conditions. Zinzinhédo and colleagues warn that no cleaning recipe wins everywhere, so modelers must match preparation to the dataset in hand.
Catalog record
| CREATOR | Mahoukpégo Luc Zinzinhédo, Mèvognon Firmin Mitchozounnou, Kolawolé Valère Salako, Romain Glèlè Kakaï |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 67721 |
| LICENSE | CC_BY |
| RIGHTS | CC BY |
| SOURCE | PLOS |