In this study, material removal rate (MRR) and surface roughness (Ra) in electrical discharge machining process have been modeled to make the process more efficient and reliable. First, adaptive neuro fuzzy inference system as one of the most used methods has been applied for prediction of material removal rate and Ra. Also a proposed method, that is, nonlinear modeling by system identification, has been applied to predict material removal rate and Ra. A group of electrical discharge machining experiments considering four input variables was conducted to collect dataset for training the predictive models. At the end, the comparison of predicted results from both approaches with experimental data shows that the new method has a much better performance than the adaptive neuro fuzzy inference system approach.