This document presents a model for AC electric arc furnace considering the highly nonlinear and time varying characteristics of this type of load. Using the nonlinear differential equation that describes the voltage – current characteristic, both voltage fluctuations in the time domain and the arc length are established assuming periodic, stochastic and chaotic variations. The model is developed using MATLAB® & SimulinkTM with parameters from a typical steel company. The model parameters are divided into two groups: deterministic and dynamic. Estimation of model parameters is accomplished using different techniques. Initially, the deterministic parameters are calibrated using three different methods: Bayesian linear regression, Tikhonov regularization and maximum likelihood. To calibrate the dynamic parameters two approaches were used: neural network and support vector machines (SVMs). The multilayer neural network is used an emulator of the electric arc furnace model. The neural network is trained using data obtained from the simulator of the electric arc furnace model implemented in SimulinkTM. Once the network is trained, the parameters of interest are obtained by solving an inverse problem. The support vector machines were used to solve regression problem in the case of multiple variables. After obtaining the multidimensional regression model of arc furnace using simulation, the model parameters of interest are obtained by solving the inverse problem taking as input the real signals. Finally, a comparison of results obtained with neural networks and SVMs is shown.