Machine learning for pyrimidine corrosion inhibitor small dataset

Abstract

Machine learning (ML) approaches have been developed to predict materials’ corrosion inhibition efficiency, particularly pyrimidine compounds. Notably, the virtual sample generation (VSG) technique enhances prediction accuracy, a novel approach for handling small datasets in this context. The random forest model, the best-performing nonlinear algorithm, showed substantial accuracy improvement based on the increase in R² value from 0.05 to 0.99 and the decrease in RMSE value from 5.60 to 0.42, after applying VSG. These results underscore the efficacy of the VSG technique in boosting the predictive performance of ML models, particularly in scenarios constrained by limited data availability.