Published January 21, 2026 | Version v1.0.0
Model Open

A critical assessment of bonding descriptors for predicting materials properties : ML models and postprocessing data

  • 1. ROR icon Friedrich Schiller University Jena
  • 2. ROR icon Federal Institute For Materials Research and Testing
  • 1. ROR icon Federal Institute For Materials Research and Testing
  • 2. ROR icon Friedrich Schiller University Jena
  • 3. Université Catholique de Louvain

Description

This Zenodo record contains the machine learning models and supporting results associated with our publication titled "A critical assessment of bonding descriptors for predicting materials properties"

 

The archive includes trained machine learning models (RandomForest/MODNet/SISSO) used in the study, saved in their original serialized formats to support reproducibility. It also contains post-processing and model interpretation outputs, including feature attribution results derived from SHAP (SHapley Additive exPlanations) and permutation feature importance (PFI) analyses.

 

In addition, the record provides correlation analysis results used to support the findings reported in the paper. Details on  training procedures, data preprocessing, and evaluation protocols are described in detail in the publication.

 

Codes to completely reproduce this data can be found on Github repository associated with this work here: DigiMatChem/paper-ml-with-lobster-descriptors

Files

README.md

Files (24.6 GB)

Name Size
md5:5764707f10ab600982e6052e511ce49c
16.8 GB Download
md5:e2c8ad11b43f0265cf51502ff7fe5b24
7.7 GB Download
md5:b60c6d43054eebc933475110fa12d01c
12.8 MB Download
md5:2719618f9547c341a75278fff3f253f3
20.5 MB Download
md5:3ca4e06dc755901413b7c469c7eb840b
72.1 MB Download
md5:f21da49c8f5d0386373909f59b35896a
4.8 MB Download
md5:392e00d1c03edc9d1014babe15241b9a
4.9 MB Download
md5:1613fe062f8e47b43419af7e2046a65b
2.3 kB Preview Download
md5:a719ff63c25f33a4e6882afcc47584be
9.4 MB Download

Additional details

Related works

Is supplement to
Preprint: 10.48550/arXiv.2602.12109 (DOI)

Funding

European Commission
MultiBonds - Understanding and designing inorganic materials properties based on two- and multicenter bonds 101161771