AUTOMATED FILLING OF CONDITIONAL PROBABILITY DATA IN A DECISION SUPPORT APPARATUS THAT USES A BAYESIAN BELIEF NETWORK

摘要

A computer implemented method is provided to expand a limited amount of input from an expert automatically to conditional probability data filling a Bayesian Belief network based decision support apparatus. The conditional probability data defines conditional probabilities of states of a particular network node as a function of vectors of state values of a set of parent nodes of the particular network node in the Bayesian Belief network. The computer implemented method comprises receiving elicited conditional probability data that defines the conditional probabilities of the state values of the particular network node for a subset of all possible vectors of combinations of state values of the parent nodes. Conditional probability data defining the conditional probabilities of the state values of the particular network node for further possible vectors of state values of the parent nodes is interpolated from the elicited conditional probability data, using an interpolation function that interpolates between the elicited conditional probability data as a function of influence factors assigned to the vectors. The influence factors are determined as a function of the ranking of the state values of the parent nodes in the possible vectors.