Parametric analysis of data-driven predictive modeling of ternary nanofluid flow in an inclined Darcy–Forchheimer boundary layer

Purpose The current study performs an investigation on thermal behavior of ternary nanofluids composed of Ag–MWCNT–GO nanoparticles, which incorporate the effects of viscous dissipation, thermal radiation, buoyancy force and convective heating along with dual slip conditions. The work also aims to develop a fast predictive framework for this highly nonlinear system. Design/methodology/approach To accurately and efficiently model this highly nonlinear system, a hybrid framework is developed that combines a conventional boundary value solver (Bvp4c) with artificial neural network trained via Levenberg–Marquardt algorithm (LMA–ANN). The numerical solutions are utilized to train two separate feedforward ANN models for temperature and velocity profiles. Findings The ANN training outcomes demonstrate strong agreement with numerical results. The temperature network (10 inputs, 1 output) reached its best validation performance at epoch 21, while the velocity network (5 inputs, 1 output) reached it at epoch 35. Five-fold cross validation further confirmed stable prediction, particularly for the temperature field, with mean squared error (MSE) of 2.37 × 10−5 and R = 0.9990, while the velocity model gave mean test MSE of 1.44 × 10−3 and R = 0.9817. These results validate the ANN's capability to reproduce the Bvp4c solutions with high accuracy, delivering a rapid and dependable predictive implementation for parametric evaluation. A targeted analysis of variance (ANOVA) -based response surface analysis further shows that radiation is the dominant positive factor affecting the local Nusselt number, while magnetic damping and porous resistance reduce it through significant coupled effects. Originality/value The study combines nonlinear boundary-layer modeling, ANN-based prediction and interaction analysis within a single framework for ternary nanofluid transport in an inclined porous radiative MHD system. The LMA–ANN technique efficiently determined these multiphysics interactions and offered an efficient tool for accurate prediction and optimization in porous media thermal systems using ternary nanofluids.