Abstract: This paper establishes a prediction model which reveals the relationship between machining parameters and the process object of wire electrical discharge machining using BP neural network method. Input of the BP network includes pulse width, pulse interval, peak current, servo voltage, workpiece thickness, and output includes machining efficiency, surface roughness. The network is trained by the samples to realize the prediction of process object. The results show that the model reflects the principle of machining process of the machine tool, realizes the prediction of the machining efficiency and surface roughness under specific conditions, and the maximum predictive error is less than 10%.