Prediction of Band Gap using Composition-Based Features with Ensemble Learning

In this paper, we utilized an ensemble learning method to study the band gap of semiconductor materials. Our models were obtained using both Light GBM and Random Forest. The experimentally determined band gaps (Eg) were used. The validation results revealed that the constructed models have a reliable level of accuracy with an R2 value equal to 0.877 for the 10-fold cross-validation best model.