In modern information warfare, it is of great significance to evaluate the threat reasonably according to the changes of the battlefield situation for realizing the automation of command system. In order to overcome the shortcomings of traditional rule algorithm, an intelligent threat assessment algorithm based on deep learning is designed. First, the principle of index selection about the warship air defense threat judgment is elaborated in the paper, and the indexes reflecting the aerial targets threat are selected from many different angles basis on this rule. Then, the warship air defense threat judgment can be attributed to a decision problem which can be solved by using the neural network model. The advantage of the method is that the output of the threat sequencing is rapid， reliable and easy to realize. It can also be updated as the constantly changing aerial threats. Finally, the validity and feasibility of the model are verified in the data sets of typical combat simulation scenarios.