Abstract: On threat assessment of marine formation air defense target, the sample data are less and easy to be missing, and existing evaluation methods rely too much on expertise and are difficult to carry out dynamic assessment on time series. In order to solve these problems, a threat assessment method based on dynamic Bayesian networks (DBN) of constraint parameter learning is proposed. The AR(p) model is used to predict the missing data on the time series, so as to obtain a complete sample of small data sets. On the basis of this, a reasonable parameter constraint model is constructed according to expertise experience. The parameter learning is carried out by Bayesian estimation under the parameter constraint model. The learning parameters are brought into DBN to get the threat evaluation results. The utility theory is introduced to sort the results of threat assessment. The simulation results show that the evaluation method is reasonable and more accurate in the condition of absence of small sample data.

Key words: dynamic Bayesian networks (DBN), constraint model, parameter learning, AR(p) model, small sample data missing, threat assessment and ranking, utility theory