HOME > Matminer

Matminer

Matminer is an open-source Python library that accelerates data-driven materials discovery. It consolidates ready-made datasets, automated data-retrieval utilities, and a comprehensive catalog of featurizers so researchers can move from raw compositions or structures to machine-learning-ready tables with minimal boilerplate code.

Information
Case / Article
Learning Content
Review
Developer’s Voice

Information

Official site	https://hackingmaterials.lbl.gov/matminer
Openness	★★★
License	Modified BSD License
Core Developers	Developers: HackingMaterials group, Lawrence Berkeley National Laboratory (LBNL), with contributions from the Materials Project and the broader materials informatics community.
Related Papers	Logan Ward et al., “Matminer: An open source toolkit for materials data mining,” Computational Materials Science 152, 60–69 (2018). doi:10.1016/j.commatsci.2018.05.018 Kristin A. Persson et al., “The Materials Project: A materials genome approach to accelerating materials innovation,” APL Materials 1, 011002 (2013). doi:10.1063/1.4812323
Other	Description: Matminer organizes the typical materials informatics workflow into modular subpackages. `matminer.datasets` exposes one-line access to 40+ curated datasets covering thermoelectrics, elasticity, dielectric constants, metallic glasses, and more—each bundled with provenance metadata. `matminer.data_retrieval` wraps external repositories such as the Materials Project, Citrination, and MPDS to build custom datasets through authenticated APIs. The `matminer.featurizers` namespace contains 70+ descriptors spanning composition statistics (Magpie), structural fingerprints (SiteStatsFingerprint, BondFractions), electronic-structure summaries (BandStructureFeatures, DOSFeaturizer), and conversion tools that bridge ASE, pymatgen, and pandas. By standardizing these tasks, Matminer lowers the barrier for MatDaCs users who want to benchmark algorithms, reproduce published datasets, or compare feature families (including DScribe’s local-environment representations) within a single pandas-centric workflow.