Surface roughness is the main indicator of technological performances of a component for electrical discharge machining (EDM). EDM process of manganese alloyed cold-work tool steel was modelled. In this paper we used the fuzzy logic (FL) and neural network (NN) to predict the effect of machining variables (discharge current and pulse duration) on the surface roughness of manganese alloyed cold-work tool steel in order to improve and increase its range of application. The experiments are carried out on manganese alloyed cold-work tool steel, processed with electrodes made of copper. The values of surface