The complex and stochastic nature of the electro-discharge machining (EDM) process has frustrated numerous attempts of physical modeling. In this paper two supervised neural networks, namely back propagation (BP), and radial basis function (RBF) have been used for modeling the process. The networks have three inputs of current (I), voltage (V) and period of pulses (T) as the independent process variables, and two outputs of material removal rate (MRR) and surface roughness (Ra) as performance characteristics. Experimental data, employed for training the networks and capabilities of the models in predicting the machining behavior have been verified. For comparison, quadratic regression model is also applied to estimate the outputs. The outputs obtained from neural and regression models are compared with experimental results, and the amounts of relative errors have been calculated. Based on these verification errors, it is shown that the radial basis function of neural network is superior in this particular case, and has the average errors of 8.11% and 5.73% in predicting MRR and Ra, respectively. Further analysis of machining process under different input conditions has been investigated and comparison results of modeling with theoretical considerations shows a good agreement, which also proves the feasibility and effectiveness of the adopted approach.