A graph convolutional network （GCN） is proposed to solve the problems such as complex target attributes and unstructured data during target threat assessment of air combat and improve the assessment efficiency. The improved criteria importance through intercriteria correlation （ICRITIC） method is introduced to build a target threat assessment model based on ICRITIC-GCN. In view of the spatial topology and attribute complexity of battlefield threat targets， the advantages of GCNs in processing non-Euclidean data are used for learning and training. To address the problem that the traditional method is too subjective in attribute weighting， the ICRITIC method considers the correlation between attributes and the amount of information on attributes and distributes attribute weights objectively. The simulations show that the algorithm is improved in processing efficiency and accuracy when solving the multi-target threat assessment problem.
