In the present study, the preparation of a novel graft yield of acrylonitrile onto cross-linked (alginate/polyvinyl alcohol) (CAP) beads was considered in the presence of potassium persulfate (KPS) as an initiator. The effects of operational parameters such as time, temperature, amount of backbone, concentration of monomer and initiator on grafting percentage were evaluated. The grafting process was modeled using the Levenberg-Marquardt algorithm of artificial neural network. The high correlation (R-2 = 0.9913) between the experimental data and model outputs showed that the Levenberg-Marquardt network is able to predict the grafting of acrylonitrile on CAP backbone accurately. The physicochemical properties of grafted samples were investigated by Fourier transform infrared, scanning electron microscopy and thermogravimetric analysis (TGA). The decomposition temperature values obtained from TGA demonstrated that the thermal stability of CAP increased as a result of grafting. Finally, the results showed that grafting of acrylonitrile onto CAP beads successfully occurred under studied experimental conditions. [GRAPHICS] .