Thermal Properties · foundry-ml
dataset_thermalcond_aflow
Thermal conductivity values and material compositions from AFLOW for machine learning screening of thermal materials.
What dataset_thermalcond_aflow is
This tabular benchmark dataset brings together 4,887 thermal conductivity records derived from the AFLOW (Automatic Flow for Materials Discovery) repository. Assembled by Conrad L. Clement, Steven K. Kauwe, and Taylor D. Sparks, it is published via Foundry-ML and the Materials Data Facility to support data-driven materials research.
The primary aim of the dataset is to enable computational screening of materials for thermal management applications. By mapping material compositions to thermal transport properties, it allows researchers in materials science, physics, and chemistry to develop predictive machine learning models for both tabular regression and classification.
How it was made
The dataset is organised into tabular records consisting of two primary fields: the chemical composition of the material (formula) acting as the input, and the thermal conductivity (target) expressed in Watts per metre-Kelvin (W/m-K).
All entries are provided in a single training split formatted in Parquet, designed for direct ingestion into analysis pipelines using Foundry-ML, Hugging Face Datasets, Pandas, or Polars.
Key facts
- Total records
- 4,887
- Input feature
- formula (Material composition)
- Target variable
- target (Thermal conductivity in W/m-K)
- Task categories
- Tabular regression, tabular classification
- Data format
- Parquet
- Originating database
- AFLOW
- Release year
- 2020
- Authors
- Conrad L. Clement, Steven K. Kauwe, Taylor D. Sparks
What you can use it for
- Thermal conductivity regression
- Training machine learning regressors to predict continuous thermal conductivity values directly from chemical composition formulae.
- Thermal management screening
- Screening large candidate composition spaces to identify materials suited for thermal insulation or heat dissipation.
- Property binning and classification
- Classifying materials into high- or low-conductivity categories using tabular classification algorithms.
Explore the data
The schema, real rows, full-text search and a SQL filter — read in your browser, nothing downloaded.
Explore
Read the real rows without downloading anything
Reading rows…
LiveRead from foundry-ml/dataset_thermalcond_aflow at the moment you asked. Nothing is cached or stored — every row above came from that request.
How the columns are distributed
Over 4,887 rows of default/train
formula
string_text- min
- 2
- median
- 8
- mean
- 7.25
- max
- 15
target
float- min
- 9.9e-3
- median
- 3.67
- mean
- 6.65
- max
- 205.13
Strengths and limits
Good at
- Provides 4,887 curated thermal conductivity data points drawn directly from the AFLOW database.
- Uses a minimalist tabular structure pairing chemical composition inputs with thermal conductivity values in standard physical units (W/m-K).
- Compatible with standard scientific Python frameworks including Pandas, Polars, and Hugging Face Datasets via Parquet.
Watch out for
- Input features are limited to composition formulae, lacking crystal structure, lattice parameters, or temperature conditions.
- Contains only a single train split, requiring researchers to manually partition data into validation and test sets.
- Excludes other thermodynamic or mechanical properties, restricting analysis purely to thermal conductivity.
Questions
Can I look inside dataset_thermalcond_aflow without downloading it?
Yes. This page reads its schema, real rows and per-column statistics in the browser, with full-text search and a SQL filter. Nothing is downloaded to your machine.
What licence is dataset_thermalcond_aflow under?
Its authors declare other. Read the terms on the original page before reusing the data.
What units are used for the thermal conductivity target?
Target thermal conductivity values are measured in Watts per metre-Kelvin (W/m-K).
What inputs are provided to predict thermal conductivity?
The dataset contains a single input feature, formula, which describes the material's chemical composition.
How many data points are included?
The dataset contains 4,887 thermal conductivity measurements.
What predefined data splits exist in the dataset?
The dataset comes structured with only a single training split (train).
How to cite it
Clement, Conrad L. and Kauwe, Steven K. and Sparks, Taylor D. Benchmark AFLOW Data Sets for Machine Learning (Thermal conductivity). Materials Data Facility, 2020. https://doi.org/10.18126/s4qn-b840
Written from dataset_thermalcond_aflow's own dataset card. Statements citing a figure that does not appear there are removed before publication. Licence and running details come from our registry.