基于整体推断的Bayes方法及其在精度评定中的应用

The Bayesian Integral Inference Method and Its Application in Accuracy Evaluation

验前分布的表示及验前信息的融合是Bayes小子样理论应用中的关键问题。根据同一型号武器试验中不同状态下信息的横向互补特性,提出一种多源信息的整体推断方法,将Dirichlet分布引入多源信息权重系数的验前信息中,建立基于Bayes网络的权重系数推断模型,利用MCMC方法更新所有节点信息,得到了合理的权重系数验后分布,解决了多源信息加权融合中权重系数难以确定的问题。仿真结果表明,该方法可以有效地融合验前分布,在精度评定中有一定的应用前景。

The expression of prior distribution and prior information fusion are greal importance in Bayesian theory＇s application. An approach about the integral inference of prior information is proposed based on one type weapon＇ s complimentary test information in different conditions. The Dirichlet distribution is introduced as the prior distribution of importance factors in multisource information and the inference model for importance factors are established by Bayesian networks. The posterior distribution can be obtained reasonably through the updated nodes by MCMC method. Hence, the problem of importance factors inference is settled. Simulation results show that this approach is able to fuse the prior distributions effectively and has a bright application prospect in accuracy evaluation.