Gaussian process regressions on hot deformation behaviors of FGH98 nickel-based powder superalloy

•The conventional Arrhenius models have been proposed to describe the hot deformation behaviors of FGH98 nickel-based powder alloys, and the Bayesian optimization technique was employed to improve the accuracy of conventional Arrhenius models. The activation energy of the FGH98 superalloy given by the Bayesian optimized Arrhenius models is 1,222,357 J mol–1.•Machine learning-based models for predicting hot deformation behaviors of FGH98 nickel-based powder alloys are developed. The machine learning-based models show better predictive and generalization ability than the conventional Arrhenius model.•The optimal parameters of the kernel function in the Gaussian processing model for predicting flow stresses are identical to those in the model for predicting peak stresses, implying a similar underlying physics in peak stresses and other flow stresses during hot deformations.