A method for estimating the effectiveness factor in a catalytic pellet submitted to deactivation using neural networks is proposed. When a catalyst is deactivated by poisoning, the function eta = eta (t,Phi) presents a minimum when strong diffusional resistances exist. In this particular case, the few methods published in the literature are not able to calculate eta. A feedforward neural network trained with the back-propagation algorithm was used to estimate the effectiveness factor. This methodology is especially useful when the function eta = eta (t,Phi) presents a minimum. The predicted values using the neural network successfully fit those obtained solving the differential equation system. An extrapolation using temperatures outside the training range can be satisfactorily performed.