Model-Based Evaluation of Long-Range Resource Allocation Plans

摘要

This report evaluated three methods for reviewing long range resource allocation plans; two based on rules with qualitative arithmetic and one based on a microeconomic model combined with quantitative reasoning and a search procedure. Evaluating a resource plan may be viewed as a process by which a reviewer compares what he knows or believes to what he is told by the numbers in a plan. For a computer program to perform such comparisons, it must use some representation of a reviewer's knowledge and beliefs. While initial representation of this knowledge in this study was in terms of if-then rules, relatively simple linear-equation models could be used to encode the same information at a more fundamental, structural level. In the study an algorithm was developed for exploring equational models, and the influence measure necessary to make quantitative, but still approximate, inferences from them. (Keywords: Expert systems; Knowledge engineering; Rule-based systems; REMUS; Logic.)