<p>Pipeline corrosion poses a major problem in the oil and gas industry, as attempts at predictive modeling face significant hurdles in providing accurate predictions of corrosion rates. Empirical and stochastic models traditionally used prove ineffective in predicting the nonlinear and multifaceted nature of corrosion processes and hence result in low accuracy when applied to actual data. In an attempt to overcome these limitations, a novel hybrid model was presented that combines Support Vector Regression (SVR) and Random Forest (RF) with optimization techniques like the Sparrow Search Algorithm (SSA), Hunger Games Search (HGS), and Chaos Game Optimization (CGO) to improve the predictive power of pipeline corrosion. The proposed model is able to fill existing gaps in methods by taking into consideration a wide range of variables such as temperature, flow velocity, internal pressure, CO2 pressure, shear stress, and the effectiveness of corrosion inhibitors. Optimized hybrid models through metaheuristic methods show a better alternative compared to traditional methods. Empirical results from a comprehensive database of corrosion show the SVR-HGS model manifests much superior performance compared to other models, displaying high accuracy in corrosion rate prediction with a testing RMSE of 0.672 and an R² value of 0.685. These findings highlight the potential of hybrid machine learning models to improve predictions in terms of precision as well as optimize maintenance strategies in the oil and gas industry.</p> Graphical abstract <p></p>