This research introduces a hybrid computational and machine learning framework for modeling improved oil recovery utilizing graphene oxide--aluminum oxide (GO–Al \(_2\) O \(_3\) ) hybrid nanofluids dispersed in vacuum residue over a stretching sheet. The physical model integrates the influences of mixed convection, magnetic field, changing thermal conductivity, and entropy generation. The controlling boundary layer equations are converted into nonlinear ordinary differential equations using similarity transformations and solved numerically using the Runge--Kutta method in conjunction with a shooting strategy. An artificial neural network (ANN) is trained on simulation results to forecast essential output parameters, including Nusselt number, skin friction coefficient, and dimensionless velocity, based on shifting physical inputs. The artificial neural network exhibits superior accuracy and generalization ability, facilitating rapid estimation of flow and heat transfer properties without the need for iterative numerical calculations. The findings validate that hybrid nanofluids augment thermal conductivity and diminish oil viscosity; hence enhancing sweep efficiency and heat transfer. The artificial neural network-integrated framework provides a practical and efficient prediction instrument for optimizing nanofluid-assisted thermal recovery operations under realistic reservoir conditions.
