An experimental and modeling investigation was carried out on removal of Methylene blue from a synthetic wastewater. Adsorption of the dye with Azolla fern was experimentally investigated in different operating conditions. Variable parameterswere initialMethyleneBlue concentration,Azolla doze, Azolla pre-treatment pH, contact time, adsorption pH and agitation rate. An artificial neural network with 6 neurons in input layer and one neuron in output layer was designed and trained to predict the removal efficiency of Methylene Blue at various conditions. Different number of neurons in the hidden layer, transfer functions and different training algorithms were examined and the optimum network was obtained by comparison of correlation coefficient and mean of square error. The experimental results showed that 8.5, 3, 150 min, 150 rpm, 1 g/lit are the optimumvalues for pH of pre-treatment, Adsorption pH, contact time, agitation rate and Azolla concentration, respectively. The investigation of modeling results showed that a network with 6, 15 and 1 neurons in input, hidden and output layers with Hyperbolic Tangent Sigmoid and linear transfer functions in hidden and output layers which is trained using Levenberg-Marquardt algorithm can predict the removal ofMethylene blue with the best precision.