Benchmarking Machine Learning Models on a Dielectric Constant Database for Bandgap Prediction

Volume 15, Issue 1 (March 2024), pp. 10–12

Download PDF

@article{jocse-15-1-2,
  author={Mohammad Hadi Yazdani and Paulo S. Branicio and Ken-ichi Nomura},
  title={Benchmarking Machine Learning Models on a Dielectric Constant Database for Bandgap Prediction},
  journal={The Journal of Computational Science Education},
  year=2024,
  month=mar,
  volume=15,
  issue=1,
  pages={10--12},
  doi={https://doi.org/10.22369/issn.2153-4136/15/1/2}
}

In this study, we investigate the performance of several regression models by utilizing a database of dielectric constants. First, the database is processed using the Matminer Python library to create features, and then divided into training, validation, and testing subsets. We evaluate several models: Linear Regression, Random Forest, Gradient Boosting, XGBoost, Support Vector Regression, and Feedforward Neural Network, with the objective of predicting the bandgap values. The results indicate superior performance of tree-based ensemble models over Linear Regression and Support Vector Regression. Additionally, a Feedforward Neural Network with two hidden layers demonstrates comparable proficiency in capturing the relationship between the features generated by Matminer and the bandgap target values.