Abstract: To reduce blind experimentation, optimize experimental designs, enable intelligent exploration, and achieve rapid prediction of material properties, data-driven machine learning methods were employed to train the four prediction models—BP neural network, support vector machine regression, random forest, and XGBoost. These methods utilized forming processes, chemical compositions, and heat treatment regimens as input features to predict the hardness and tensile strength of the selective laser melted nickel-based superalloy materials. The results indicate that the XGBoost algorithm prediction model achieves the best fitting performance. The XGBoost prediction model was optimized using Bayesian optimization algorithm, the ASBO-XGBoost prediction model of nickel-based superalloys was established. The determination coefficient R² of training set and testing set of the hardness prediction model achieves 0.961 and 0.953, respectively, while the R² of training set and testing set of the tensile strength prediction model achieves 0.993 and 0.985, respectively, demonstrating excellent fitting performance. To evaluate the model's predictive performance, experimental validation was conducted by varying process parameters, heat treatment regimens and alloying element contents. The absolute mean errors between the model's predictions and experimental results for hardness and tensile strength are 28.7 HV and 20.4 MPa, respectively. This confirms the accuracy and reliability of the ASBO- XGBoost model in predicting the strength and hardness of the nickel-based superalloys.