Wire-EDM is a highly complex process, which is characterized by non-linear behavior. Due to a large number of input parameters, a data mining approach based on machine learning is followed in this paper to model this process. The model was trained on experimental data collected from carefully conducted experiments. The model was also tested on additional data. It is found that the model built using data mining provides results with desired accuracy. The important parameters have been identified and reported. © 2003 Elsevier Science B.V. All rights reserved.