SAIP2026

DEVELOPMENT OF MACHINE LEARNING MODELS TO PREDICT PROPERTIES OF HALIDE PEROVSKITE MATERIALS

8 Jul 2026, 16:20

1h 20m

Great Hall ( University of the Western Cape)

Poster Presentation Track A - Physics of Condensed Matter and Materials Poster Session 2

Speaker

Ms Angela Mmasefako Maboa (University of Limpopo)

Description

Angela Maboa1, Keletso Monareng1, Petros Ntoahae1 and Rapela Maphanga2

1Department of Physics, University of Limpopo, Private bag X 1106, Sovenga, 0727, Polokwane, South Afri-ca
2Renewable and Sustainable Energy Research Centre, Sol Plaatje University, Private Bag X 5008, Kimberly, 8300, South Africa
3 National Institute for Theoretical and Computational Sciences, NITheCS, Gauteng, 2000, South Africa

Abstract
Halide perovskite materials are the foremost candidates for the next generation of optoelec-tronics, but the vast chemical space makes experimental screening for the stable phases time-consuming and expensive. This study presents the development of robust supervised machine learning models to predict the four key properties of halide perovskite materials, namely : energy,band gap, formation energy, energy per atom, and Fermi energy. The data set sourced from the Material Project Database was curated and engineered to extract meaningful descriptors using structural, compositional, and electronic features. Several ensemble-based models, including Random Forest, Extra Trees, and Gradient Boosting, were evaluated using the two metrics, regression score (R²) and mean square error (MSE). The analysis revealed that the Random Forest model was the most effective in capturing electronic transitions, yielding the highest precision for band gap predictions (R² = 0.989, MSE = 0.001). The Extra Tree Regressor outperformed others in predicting the formation energy (R² = 0.831, MSE = 0.100), the energy per atom (R² = 0.788, MSE = 1.500) and the Fermi energy (R² = 0.830, MSE = 1.041). The findings validate the use of ensemble learning as a robust approach for mapping the intricate relationships between the structure and properties of halide perovskites. This work provides a scalable computational bridge for accelerating the deployment of efficient halide perovskites for clean energy.