Abstract
This paper presents a quantitative structure–property relationship (QSPR)-based machine learning (ML) framework designed for predicting corrosion inhibition efficiency (CIE) values in natural organic inhibitor compounds. The modeling dataset comprises 50 natural organic compounds, with 11 quantum chemical properties (QCP) serving as input features, and the target variable being the corrosion inhibition efficiency (CIE) value. To enhance the predictive accuracy of the ML model, the kernel density estimation (KDE) function is employed to generate virtual samples during the training process, with the overarching goal of refining the precision of the ML model. Three distinct models, namely random forest (RF), gradient boosting (GB), and k-nearest neighbor (KNN), are tested in the study. The results demonstrate a noteworthy enhancement in the prediction performance of the models, attributable to the incorporation of virtual samples that effectively improve the correlation between input features and target values. Consequently, the accuracy of the predicted CIE values is significantly augmented, aligning more closely with the actual CIE values. Performance improvements were evident across all models after the incorporation of virtual samples. The GB, RF, and KNN models exhibited increments in R2 values from 0.557 to 0.996, 0.522 to 0.999, and 0.415 to 0.994, respectively, concomitant with the introduction of 500 virtual samples. Additionally, each model demonstrated a notable reduction in RMSE values, transitioning from 1.41 to 0.19, 1.27 to 0.10, and 1.22 to 0.16, respectively. While the GB model initially outperformed others before the addition of virtual samples, the performance of the model exhibited fluctuation as the number of virtual samples varied. This behavior suggests that the KDE function provides a certain level of resilience against model variations. The proposed approach contributes to the effective design and exploration of corrosion inhibitor candidates, offering a reliable and accurate predictive tool that bridges the gap between theoretical studies and experimental synthesis.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.