Accurate measurement of ultrasonic velocities is the essential part of structural characterization of materials. The longitudinal and shear ultrasonic velocities in multicomponent glass systems can be measured experimentally by the conventional pulse-echo technique which needs highly sophisticated instrumentation and so costly. On the theoretical evaluation side the usual statistical simple or multiple regression analysis do not work well to predict the velocities, since the relationship between the characteristic parameters of the components and the ultrasonic velocities are highly non-linear and quite complex. In situations like this artificial intelligence techniques are the best choice to solve the problem. Present work deals with the development of a multiplayer perceptron (MLP) artificial neural network (ANN) to predict the ultrasonic velocities in binary oxide glass systems.