Comparison of Linear and Non-Linear Machine Learning Algortima for Predicting the Effectiveness of Plant Extracts as Corrosion Inhibitors

Abstract

This research aims to develop a Machine Learning (ML) model that can predict the corrosion inhibitor potential of plant extracts with high accuracy. Corrosion is a serious problem in industry because it can reduce the service life of materials and cause economic losses. This research focuses on the use of green inhibitors, especially plant extracts, which are considered environmentally friendly and have high anticorrosion efficiency. The dataset used includes molecular and physicochemical features of plant extracts. The ML model development process involves data normalization, selection of linear and non-linear ML algorithms, model training with k-fold crossvalidation, and model performance evaluation using regression metrics such as MSE, RMSE, MAE, and R2. Experiments compare various ML algorithms and show that the AdaBoost Regressor (ABR) model exhibits the best prediction performance with the highest R2 value of 0.993 and a low MSE of 0.002. These results provide new insights into the potential of ML models to predict effective corrosion inhibitors from plant extracts. The ABR model had a low prediction error, indicating high accuracy in predicting corrosion inhibition efficiency. In addition, the analysis of important features shows that two features, Conc and LUMO, have a significant influence on the ABR model. This research makes an important contribution to the development of effective prediction methods in the corrosion control industry. The ABR model can serve as a basis for designing more effective and environmentally friendly corrosion inhibitor materials, as well as a reference for other researchers in developing ML models that accurately predict the corrosion inhibition efficiency of plant extracts.