In this study graphene oxide (GO) was functionalized using amine and used as inorganic filler for preparation of mixed matrix membranes (MMMs) using polyethersulfone (PES) as polymer matrix. These membranes were applied for separation of CO2 from CH4. The effects of filler loading, feed temperature and feed pressure on CO2/CH4 selectivity of the MMMs were investigated. The results indicated that addition of amine-functional graphene oxide in the casting solution enhanced the membrane gas permeance and CO2/CH4 ideal selectivity. SEM images and FTIR analysis were used to characterize the filler particles and the synthesized membranes. SEM images also indicated that, there were appropriate distribution particles in the polymer matrix. Among different types of artificial neural networks (ANN), radial basis function (RBF) network was used to model performance of the MMMs. For training of the RBF model, 70% of the collected experimental data was used and the model was tested using the rest 30% data. The mean square error (MSE) and correlation coefficient (R) were used for investigating performance of the RBF model. The results showed that the RBF model is suitable and efficient for predicting performance of the PES/amine-functional graphene oxide (AFGO) MMMs.