BandGap-ml 0.3.3

Project for predicting band gaps of inorganic materials by using ML models

BandGap-ml v0.3.3

Table of Contents

• Project Description
• Prepare Workspace Environment with Conda
• Models Construction
• Usage
• Author
• License

Project Description

Project for predicting band gaps of inorganic materials by using ML models.

Try out new Frontend Web Interface running at:

https://bandgap-ml.streamlit.app/

Prepare Python Workspace Environment with Conda

1. Download Miniforge for Unix-like platforms (macOS & Linux)

# Download the installer using curl or wget or your favorite program:
curl -L -O "https://github.com/conda-forge/miniforge/releases/latest/download/Miniforge3-$(uname)-$(uname -m).sh"

# OR 
wget "https://github.com/conda-forge/miniforge/releases/latest/download/Miniforge3-$(uname)-$(uname -m).sh"

# Run the script:
bash Miniforge3-$(uname)-$(uname -m).sh

and follow instructions. See the documentation for Miniforge for more information.

# 2. Create and activate the conda environment
conda create --name bandgap-ml "python<3.13"
conda activate bandgap-ml

# 3. Install BandGap-ml
# 3.1 From PyPI
pip install BandGap-ml

# 3.2 Or install the latest version from the GitHub repository
pip install git+https://github.com/alexey-krasnov/BandGap-ml.git

# 3.3 Or install the latest version in editable mode from the GitHub repository
git clone https://github.com/alexey-krasnov/BandGap-ml.git
cd BandGap-ml
pip install -e .

• Where -e means "editable" mode.

Data source

For training Random Forest Classifier and Regression models, we adopted data provided in the following paper:

• Zhuo. Y, Mansouri Tehrani., and Brgoch. J, Predicting the band gaps of inorganic solids by machine learning, J. Phys. Chem. Lett. 2018, 9, 1668-1673.

Models construction

To perform model training, validation, and testing, as well as saving your trained model, run the following command in the CLI:

python band_gap_ml/model_training.py

This command executes the training and evaluation of RandomForestClassifier and RandomForestRegressor models using the predefined paths in the module.

Usage

We provide several options to use the package

1. Jupyter Notebook file:

Jupyter Notebook file in the notebooks directory provides an easy-to-use interface for training models and use them for Band Gap predictions.

2. Use the package inside your Python Code:

Train models

from band_gap_ml.model_training import train_and_save_models

train_and_save_models()

Make predictions of Band Gaps by using the BandGapPredictor class:

from band_gap_ml.band_gap_predictor import BandGapPredictor

# Initialize the predictor with default best model
predictor = BandGapPredictor()

# Or specify a different model type and path to the model
# predictor = BandGapPredictor(model_type='RandomForest', model_dir= <YOUR_PATH_TO_THE_MODEL>)
# predictor = BandGapPredictor(model_type='GradientBoosting')
# predictor = BandGapPredictor(model_type='XGBoost')

# Prediction from csv file containing chemical formulas
input_file = 'samples/to_predict.csv'
predictions_df = predictor.predict_from_file(input_file)
print(predictions_df)

# Prediction from one or multiple chemical formulas
formula_1 = 'BaLa2In2O7'
formula_2 = 'TiO2'
formula_3 = 'Bi4Ti3O12'

# Single formula prediction
single_prediction = predictor.predict_from_formula(formula_1)
print(single_prediction)

# Multiple formulas prediction
multiple_predictions = predictor.predict_from_formula([formula_1, formula_2, formula_3])
print(multiple_predictions)

# Save predictions to a CSV file
multiple_predictions.to_csv('predictions_results.csv', index=False)

3. Use frontend web interface

• Go to https://bandgap-ml.streamlit.app/ to check out the web interface
• Or run web interface on your local machine. In CLI run the command:

streamlit run frontend/band_gap_ml_app.py --server.address=0.0.0.0 --server.port=5005

The command will refer you to the BandGap-ml user web interface.

Author

Dr. Aleksei Krasnov [e-mail]

Citation

• Zhuo. Y, Mansouri Tehrani., and Brgoch. J, Predicting the band gaps of inorganic solids by machine learning, J. Phys. Chem. Lett. 2018, 9, 1668-1673. https://doi.org/10.1021/acs.jpclett.8b00124

License

This project is licensed under the MIT - see the LICENSE.md file for details.