<p>Corrosion is a major degradation mechanism that significantly reduces the service life of metallic materials and leads to substantial economic losses in various industries. Therefore, it is of great importance to comprehensively address both technical and economic consequences of corrosion, to increase material strength and to encourage sustainable engineering practices. In this context, reliable and efficient predictive modeling of corrosion behavior of high-entropy alloys (HEAs) presents a high-dimensional regression problem due to the inclusion of numerous alloy components, empirical parameters, and environmental variables. This can lead to the inclusion of unnecessary or low-contributing variables in machine learning models, negatively impacting generalization performance. Therefore, this study proposes a machine learning framework based on a feature selection approach using the Cheetah Optimization Algorithm (COA) for predicting the corrosion rate of HEAs. XGBoost, Random Forest, Extra Trees, Gradient Boosting, AdaBoost, and Decision Tree regression models were trained using selected feature subsets and evaluated with a 10-fold cross-validation method. Analyses were performed separately for the post-experiment scenario, where corrosion current density was included in the model, and the pre-experiment scenario, where corrosion current density was excluded. The results show that optimized feature selection significantly contributes to model accuracy, generalizability, and physical interpretability in modeling the corrosion behavior of HEAs.</p>