starrydata-utils

Python utility libraries for analyzing thermoelectric and magnetic materials data from the Starrydata database.

Overview

• starrydata_utils.py — Thermoelectric materials: composition parsing, data loading, spline interpolation, derived property calculation (power factor, ZT, weighted mobility), material family classification (~50 families), PCA/clustering, and plotting (matplotlib + plotly).
• starrydata_magnetic_utils.py — Magnetic materials: data loading, family classification, hysteresis plotting, clustering (PCA/t-SNE/NMF), and the Brillouin function. Imports shared functions from starrydata_utils.py.
• generate_data.py — Data conversion pipeline: reads raw Starrydata CSV files, performs spline interpolation and derived property calculations, and outputs processed CSV files.
• build_notebook.py — Generates a Colab notebook from starrydata_utils.py by splitting at section markers.

Quick start

pip install -r requirements.txt

Download and process the dataset

1. Download the latest Starrydata dataset from Google Drive and place it in data/raw/:

data/raw/starrydata_dataset_YYMMDD/
├── starrydata_curves.csv
├── starrydata_samples.csv
└── starrydata_papers.csv

2. Update the RAW_DIR path in generate_data.py if needed, then run:

python generate_data.py

This produces processed files in data/processed/.

Use as a library

from starrydata_utils import (
    download_dataset, load_curves, load_samples,
    spline_interpolate_curves, classify_material_families,
    calculate_derived_properties, TEplot4, single_plot,
)

# Download dataset
datapath = download_dataset('1py40fDLkTW2kcGx-ie7xHxG2Iqisfcuk')

# Load data
df_curves = load_curves(datapath)
df_samples = load_samples(datapath)

# Classify material families
df_samples = classify_material_families(df_samples)

Generate Colab notebook

python build_notebook.py

Data flow

download_dataset() → load_curves() + load_samples()
    → spline_interpolate_curves() at 100K intervals
    → classify_material_families()
    → calculate_derived_properties()
    → TEplot / TEplot4 / single_plot / plotly_2d / ...

Three core DataFrames: df_sample (metadata + composition), df_curve (raw curves), df_int (interpolated + derived properties).

Dependencies

• pandas, numpy, scipy
• pymatgen (composition parsing)
• matplotlib, plotly (visualization)
• scikit-learn (PCA, KMeans, t-SNE, NMF)
• gdown (Google Drive download)
• tqdm (progress bars)

Links

• Starrydata