<p>This study presents an integrated approach of finite element simulation of extrusion and machine learning to predict the peak load required for warm extrusion of a biocompatible Mg–Zn–Ca alloy. The FE simulations of extrusion were carried out for three different alloy compositions using five different flow stress models, wherein design and process variables such as half-die angle, billet temperature, and ram speed were varied to minimize the extrusion peak load. The simulation data was utilized in a machine learning approach for the training and testing of a gradient boosting regressor model. The accuracy of the machine learning model was improved by tuning of hyperparameters such as n_estimators, learning_rate, max_depth and min_samples_split applying GridSearchCV along with five—fold cross validation. The developed gradient boosting regressor model demonstrated high accuracy on the training dataset, achieving an R² value of 0.99, a mean absolute error of 0.46, and a mean square error of 0.50. The score of feature importance indicates that the billet temperature and half-die angle are influential parameters for predicting extrusion peak load. To validate the gradient boosting regressor model, extrusions of Mg–Zn–Ca alloy were conducted in the laboratory at three different temperatures (300&#xa0;°C, 350&#xa0;°C, and 400&#xa0;°C) with three different half die angles (15°, 30°, 45°). The predicted loads demonstrated a reasonably close agreement with the experimental results, validating the utility of this hybrid approach.</p> Graphical Abstract <p></p>