Accelerating band gap prediction for solar materials using feature selection and regression techniques

We present a novel approach to apply machine learning techniques to build a more robust prediction model for band-gap energies (BG-E) of chalcopyrites, a class of materials for energy applications in the fields of solar energy, photocatalysis, and thermoelectrics. Guided by knowledge from domain experts and by previous works on the field, we aim to accelerate the discovery of new solar materials. Our objectives are two folds: (i) Identify the optimal set of features that best describes a given predicted variable. (ii) Boost prediction accuracy via applying various regression algorithms. Ordinary Least Square, Partial Least Square and Lasso regressions, combined with well adjusted feature selection techniques are applied and tested to predict the band gap energy of chalcopyrites materials. Compared to the results reported in Zeng et al. (2002), Suh et al. (1999, 2004), and Dey et al. (2014), our approach shows that learning and using only a subset of relevant features can improve the prediction accuracy by about 40%.