Analysis of Information Value in an Influence Diagram

张红兵,赵杰煜,罗雪山
DOI: https://doi.org/10.3321/j.issn:1001-506x.2004.04.006
2004-01-01
Abstract:The qualitative relationship about the informational relevance of variables in graphical decision models is derived based on the research of the topology of the models and the analysis of dominance relations for the expected value of information on chance variables in terms of their positions and relationships in influence diagrams. A new method to sort the chance variables in an influence diagram with the expected value perfect information (EVPI) is introduced. The qualitative relationship can be employed to generate the non-numerical procedures for prioritizing uncertain variables in a decision model with their informational relevance.
What problem does this paper attempt to address?