运用贝叶斯网络技术对作战重心(Center of Gravity,COG)建模时,子节点条件概率表中的概率分布数量随着父节点数目的增加呈指数增长,这对担负概率估算的领域专家而言是一个巨大的挑战.分析了领域专家在估算条件概率时的启发式思维,提出一致性父节点配置的概念,并把在该配置下估算获得的条件概率和父节点的相对权重作为输人,使用加权和的方法生成其余的条件概率.该方法减少了领域专家在估算条件概率时的认知量,有利于保持概率分布的一致性.%The number of probability distributions required in the conditional probability table (CPT) of a child-node grows eXponentially with the number of its parent-nodes in the process of COG modeling with Bayesian networks technology, which is a big challenge for domain expert who is assigned to estimate these probabilities. The heuristic thought of domain expert in conditional probability estimation is analyzed, and the concept of compatible parental configurations is presented in this paper. The rest conditional probabilities in the CPT are generated by the method of weighted sum algorithm with the input of conditional probabilities estimated in the compatible parental configuration and relative weights of parent-nodes. The extent of knowledge acquisition is reduced radically when estimating conditional probabilities using this method as well as making for keeping compatibility of probability distributions.
展开▼
