Prediction of Anti-Corrosion performance of new triazole derivatives via Machine learning

• Published in: Computational and theoretical chemistry 2024-06, Vol.1236, p.114599, Article 114599
• Main Authors: Akrom, Muhamad, Rustad, Supriadi, Kresno Dipojono, Hermawan
• Format: Article
• Language: English
• Subjects: Corrosion inhibition; Machine learning; N-heterocyclic; QSPR; Triazole
• ISSN: 2210-271X
• DOI: 10.1016/j.comptc.2024.114599

[Display omitted] •XGBoost emerged as the best predictor.•ML approach shows high CIE values for Triazole compounds.•DFT calculations assume a pivotal role within the QSPR model. This paper endeavors to present an in-depth investigation into the corrosion inhibition efficiency (CIE) of novel triazole derivatives serving as corrosion inhibitors. Among the array of models considered, the extreme gradient boosting (XGBoost) model emerged as the most adept predictor in forecasting the CIE of N-heterocyclic organic compounds. This resolute preference for the XGBoost model was consistently upheld when employed in the prediction of the CIE for three newly synthesized triazole derivatives, namely, 2-[5-Phenyl-1-(2′-furanylmethylene)imino-(1,3,4)homotriazole]thio-N-(2′-furanyl)hypomethylacetylhydrazine, 2-[5-(3′-Methyl)Phenyl-1-(2′-furanylidine)imino-(1,3,4)homotriazole]thio-N-(2′-furanylidine) acetylhydrazine, and 2-[5-(4′-Methyl)Phenyl-1-(2′-furanylidine)imino-(1,3,4)homotriazole]thio-N-(2′-furanylidine) acetylhydrazine. Remarkably, this application of the XGBoost model yielded notably elevated CIE values, spanning from 88.35 % to 93.41 %. Supplementary density functional theory (DFT) calculations for these derivative compounds further substantiated the predictive trends observed through machine learning and experimental predictions. These calculations revealed robust adsorption energies falling within the range of −17.95 to −19.76 eV, in alignment with the CIE trends established through the machine learning framework.