<p>Mechanical performance of 3D-printed architected materials are highly sensitive to manufacturing process parameters, but traditional methods usually rely on fabricating representative specimens and conducting experiments to caliber the response sensitivity. In this study, a machine learning-based framework is proposed that can directly predict the mechanical performance of a classic type of architected lattice, while also enabling reverse design of suitable process parameters. The Latin hypercube sampling method is employed for uniform sampling within the design space, combined with Gaussian random noise techniques to improve prediction accuracy under small sample constraints. By incorporating data across the entire range of printing speeds, the model demonstrates improved robustness. Comparative results show that CatBoost outperforms other algorithms (XGBoost, LightGBM, Random Forest, and SVR) in predicting the mechanical performance of architected lattices, achieving an R² value of 0.93 and significant lower errors (MSE, RMSE, MAE). Furthermore, explainable AI tools such as SHAP values and partial dependence plots reveal the nonlinear influences of printing speed, nozzle temperature, and layer height on mechanical performance. Finally, the reverse design method successfully generates new process parameters that meet target industrial specifications. This work presents a more efficient and rapid predictive tool for the field of additive manufacturing., thereby reducing the need for costly and time-intensive fabrication and testing procedures.</p>