Comparative machine learning approach on Maxwell tetra hybrid nanofluid with porous and bioconvection over 3d...

This present study investigates the 3D magnetized Maxwell tetra-hybrid nanofluid with porous medium, viscous dissipation, joule heating, Soret effect and bioconvection over a rotating stretching sheet. Titanium dioxide $\left(Ti{O}_2\right)$ , Gold $\left(Au\right)$ , magnetic iron oxide $\left(F{e}_3{O}_4\right)$ and silicon dioxide $\left(Si{O}_2\right)$ acts as a promising multifunctional platform for biomedical transport processes. This combination has potential applications in targeted drug delivery, cancer hyperthermia, biomedical cooling and diagnostic technologies. The governing equations are converted into a set of ordinary differential equations using similarity transformations and solved by 5th order Runge-Kutta Fehlberg method along with the shooting technique. Furthermore, artificial neural network (ANN) and tree-based ensemble regression models were developed using the numerical data generated by 5th order Runge-Kutta-Fehlberg shooting technique to provide fast and accurate prediction of the transport characteristics while reducing the need for repeated numerical simulations. The proposed study provides accurate and computationally efficient prediction of transport characteristics and has potential applications in thermal management, biomedical engineering and energy systems. From the Random forest method, the average Root Mean Squared Error of Sherwood number is 0.008635 ± 0.001430. From the ANN, trainbr method gives the best performance is $2.2982\times {10}^{-9}$ at epoch 800 and trainlm method gives the best validation performance is 0.00012547 at epoch 12.