基于模糊影响图的海洋结构风险分析方法研究_英文_.pdfVIP

No.2 KANG Haigui et al.: A Fuzzy Probabilistic Influence Diagram Method to Assess … 23 indirect causality. The absence of an arc between two nodes does signify that they are conditionally independent. At the numerical level, for each node, we need to specify a set of probability values and a set of conditional probability distribution for it given the values of its predecessors. In this paper, P (A) and P (A ) is a probability that represents the degree of belief in A and A (A ’s negation). To demonstrate a probabilistic influence diagram graphical representation, consider the following example (Fig.1), where A is the parent of B and C, which are the parents of D. B and C are referred to as the children of A. C is also the parent of E. Proposition A is referred to as the source proposition, and the propositions D and E are referred to as the sink propositions. Predictive or causal inference involves reasoning from evidence about source propositions down through the diagram in the direction of links to sink propositions. It is assumed that all influence is captured in conditional probability relations between a child proposition and its parents. In other words, the child is conditionally independent of all other propositions in the network apart from its parents. If this is the case, then for proposition E, P (E ) P (E | C )P (C ) =+P (E | C )P (C ) ( 1 ) The relationship between C and E is a feature of the structure of the inference problem. For example, C may be a necessary condition for E, in which the case P (E | C ) = 0 and P (E | C ) ≤ 1, or C may be a sufficient condition for E, in which case P(E | C ) = 1 and P (E | C ) ≤ 1. Probabilitie