In the context of optimization problems, metamodels can effectively evaluate objective and constraint functions, thereby expediting the optimization process. Metamodels are typically categorized into two types: classification and regression. Each type has unique characteristics that make them suitable for different applications in optimization models. A regression model predicts the relationship between dependent and independent variables by fitting statistical equations to observed data. In contrast, a classification model assigns input data to predefined categories or groups based on patterns recognized from labeled training data. This research focuses on building and evaluating the performance of two types of metamodels—classification and regression—when integrated with an enhanced Differential Evolution (DE) algorithm to solve the nonlinear truss optimization problem. The findings indicate that classification models deliver superior optimum results and greater sta-bility in comparison to regression models. Moreover, the error rate of classification models is notably lower than that of regression models, and they also require less computation time.