Download file (7.84 MB)
Download as zip (7.02 MB)

Collection

Citation

Description:

In this study, we addressed the data scarcity problem in materials science by constructing a comprehensive dataset of dielectric materials using the Starrydata2 web system, collecting experimental data from over 20,000 samples. Using this dataset, we developed high-performance machine learning models and identified important descriptors through recursive feature elimination. As the models functioned as black boxes, we employed dimensionality reduction and clustering techniques to visualize trends in dielectric properties. By combining key factors with material clustering, we visualized the relationship between crystal lattice and dielectric permittivity in ABO3 systems, revealing a nearly linear relationship. These analyses provide an important foundation for data-driven materials research.

Rights:

• Creative Commons BY-NC Attribution-NonCommercial 4.0 International
  

Keyword: Materials data analysis, database, data curation, dielectric materials, ferroelectric, machine learning, dimensionality reduction

Date published: 2025-12-31

Publisher: Informa UK Limited

Journal:

• Science and Technology of Advanced Materials: Methods (ISSN: 27660400) vol. 5 issue. 1

Funding:

• the Japan Science and Technology Agency JPMJCR19J1 (新規結晶の大規模探索に基づく革新的機能材料の開発)

Manuscript type: Publisher's version (Version of record)

MDR DOI:

First published URL: https://doi.org/10.1080/27660400.2025.2485018

Related item:

Other identifier(s):

Contact agent:

Updated at: 2025-06-12 16:30:20 +0900

Published on MDR: 2025-06-12 16:25:29 +0900