<![CDATA[Corrosion degrades and weakens metal surfaces, leading to structural failure and significant safety hazards across various sectors. Data driven machine learning offers a rapid, cost-effective alternative to the expensive and time consuming traditional experimental methods by predicting inhibitor performance computationally. This study addresses the challenge of accurately predicting corrosion inhibition efficiency (CIE) of ionic liquid compounds. Integration of a polynomial function, especially in higher degrees, inevitably grows the dimensionality and escalates multicollinearity, but it captures deeper nonlinear interactions that the original variables alone would miss. To counterbalance this curse of dimensionality, Partial Least Squares (PLS) Regression was applied after polynomial integration to project the high-dimensional variables into a smaller set of predictors. Besides PLS, Gradient Boosting Regressor (GBR) and Support Vector Regressor (SVR) models were also developed to establish baseline performance. Although these polynomial integrated models outperformed their baseline version, the Polynomial Integrated PLS outperformed their predictive performance, yielding R2, RMSE, and MAPE of 0.73, 4.730, and 3.73%, respectively. The result of this study highlights that the integration of a polynomial function can improve the predictive performance of PLS for corrosion inhibitors.]]>
Copyrights © 2025
