The present study is proposed to develop the Adaptive Neuro-Fuzzy Inference System optimized by genetic algorithm to estimate CO2 value in permeate stream using a hollow fiber polymeric membrane for separation of binary gas containing CO2 and CH4 in natural gas. To that end, a number of 65 samples was gathered from the literature. Results indicated that the proposed ANFIS model has great potential with high correlation (R-2 = 0.9993) and less error (RMSE = 0.0064) for estimation of aforementioned parameter.