Membrane processes are inherently complex due to the uncertain relationship between membrane performance and structure, feed characteristics, the interaction between feed and membrane, and operating conditions. In the present study, a feed-forward artificial neural network (ANN) and an adaptive neuro-fuzzy inference system (ANFIS) as the powerful tool for modeling of complex and non-linear membrane characteristics in terms of permeate flux have been utilized for polysulfone/starch composite membrane. The parameters used for modeling the permeate flux behavior consist of three inputs (filtration time, trans-membrane pressure, and concentration) and the experimental permeate flux as the output. For ANN, ten neural network training functions with three-layer multi-perceptron structure whose number of neurons in the hidden layer ranging from 6 to 12 was used whereas the first type Gaussian membership function was used for input variables, and hybrid algorithm was selected for the input–output data in ANFIS model. The results of ANN and ANFIS models have shown that predicted and experimental values give good values of coefficient of determination (R2 > 0.999) and low values of mean squared error (MSE), root mean square error (RMSE), and mean absolute percentage error (MAPE) for both approaches.