<p>This study presents a machine learning-based framework to improve the aerodynamic prediction accuracy of Missile DATCOM by developing surrogate regression models for carryover interference factors and the center of pressure using wind-tunnel data. The dataset, obtained for a missile body-fin configuration with twelve fin geometries, was refined by identifying stall regions, correcting low-angle-of-attack anomalies, and removing outliers to ensure physical consistency. To model the nonlinear aerodynamic characteristics, two base learners – Multi-layer perceptron and Random Forest – were employed and combined using a stacking ensemble method. Hyperparameter optimization was conducted using Optuna, which significantly improved model generalization. The proposed ensemble models demonstrated better agreement with experimental results compared with the empirical and theoretical formulations of Missile DATCOM. Furthermore, the trained neural networks were converted into ONNX format and integrated directly into Missile DATCOM routines, enabling real-time inference and correction of interference coefficients and center-of-pressure locations. This approach provides a foundation for extending machine learning-assisted aerodynamic modeling to broader missile and aircraft configurations in future research.</p>