Abstract: First, a decision method of Beyond Visual Range(BVR) multi target air combat cooperative prioritization algorithm is presented, by which the target distribution of cooperatively attacking multiple targets by multiple fighters can be given. Then, for 2 via 4 air combat, the algorithm is realized by three layers of B P neural network, and the object attacking sequence is realized by a Self Organ Feature Mapping(SOFM) one. In accordance with the concrete features of the learning algorithm of them, appropriate sample space, training and testing methods are designed. Thus, favorable results have been obtained when they are trained and tested. This research shows that it is easy, accurate and fast to realize the object attacking sequence of air combat by means of SOFM neural network. Because they have strong capability of environment adaptation, error tolerance and real time characteristic, the realization must be helpful to the pilot’s air combat decision in a complicated and dynamic environment, and can enhance the fighting effectiveness of the aeronautical weapon system.