SMILES-based machine learning enables the prediction of corrosion inhibition capacity

Abstract

This study explores the efficacy of using a simplified molecular input line entry system (SMILES) as the sole feature, replacing quantum chemical properties (QCP), in predicting corrosion inhibition efficiency (CIE) for N-heterocyclic compounds. The gradient boosting regressor (GBR) model outperforms k-nearest neighbors (KNN), support vector regression (SVR), and other models. SMILES accurately predicts CIE for various datasets, demonstrating potential as a standalone feature. Results indicate a moderate correlation between SMILES representation and corrosion inhibition properties. The proposed method identifies novel N-heterocyclic derivatives with high CIE, suggesting its utility in discovering corrosion inhibitors.

Graphical abstract