This study develops an uncertainty aware and predictive model for the electromagnetohydrodynamic (EMHD) boundary layer flow of an ethylene glycol water based hybrid nanofluid comprising cobalt ferrite ( \({CoFe}_{2}{O}_{4}\) ) and ferric oxide ( \({Fe}_{3}{O}_{4}\) ) nanoparticles. The flow is analyzed over a permeable plate, precisely accounting for Thompson and Troian slip conditions, heat absorption/generation and suction/injection effects. Volumetric uncertainties in nanoparticles ( \({\upphi }_{1}\) , \({\upphi }_{2}\) ) are modeled as triangular fuzzy numbers (TFNs), converting the governing system into fuzzy differential form via the α-cut method which is further treated numerically via bvp4c solver in MATLAB. The fuzzy analysis captures the variability in velocity and temperature fields, with deviations between crisp and fuzzy outcomes below 0.01%, confirming high model stability. The numerical data obtained for engineering parameters serve as input datasets for training the Artificial Neural Network (ANN) algorithms, Bayesian Regularization (BR) and Levenberg–Marquardt (LM) in comparative manner, the ANN effectively provides the prediction of system’s thermal and hydrodynamic performance and validates the numerical trends with high accuracy. The Bayesian Regularization (BR) achieved a mean squared error of 2.54 × 10⁻1⁰ with R = 1, validating its reliability for predicting thermal and hydrodynamic parameters. Results illustrate that rising the electric field parameter (ε), Hartmann number (Ha) and Darcy number (K) lead to an increase in both the skin friction coefficient and Nusselt number as K rises from 0.2 to 0.8, skin friction increases from 0.7897 to 0.9864 and the Nusselt number from 0.7234 to 0.8125. Conversely, higher Eckert number (Ec) and heat generation parameter (Q) reduce thermal efficiency. This integrated fuzzy–ANN framework provides a basis for uncertainty quantification and optimized design of hybrid nanofluid based thermal systems.
