Threat evaluation is a time sensitive, uncertain and dynamic process. In order to access a profound understanding, a detailed description of 18 threat cues essential in the assessment process is given firstly. And Gaussian function is used as fuzzy membership function to deal with data required. At the same time, Bayesian network structure is designed for complex threat factors. Then a two stage evaluation structure based on the above description is proposed: Important information unknown about distant target is speculated based on the association rules and then average linear weighted method is used on the parameters to get the assessment results; Bayesian network is used when information is complete. Finally simulation results verify the validity and rationality of the model.