Aiming at the problem that the existing efficiency prediction methods are difficult to reflect the actual effectiveness of anti-missile equipment system， a method of efficiency evaluation of anti-missile equipment system based on "data-driven + deep learning" is proposed. On the basis of a large number of experimental data extraction， disposal and analysis， we construct grey wolf optimization （GWO）-deep belief network（DBN） model to train the data， so as to obtain the nonlinear fitting of the anti-missile equipment system efficiency. We conduct a simulation experiment with an anti-missile system efficiency evaluation as an example， and the results show that the evaluation method is feasible and reliable. It can provide high reference value and significance for the demonstration and improvement of the anti-missile equipment system.