Expert system models for inference with imperfect knowledge: A comparative study (Q1110349)

This paper presents a detailed comparative study of six major, leading methods for reasoning based on imperfect knowledge: (1) Bayes' rule, (2) Dempster-Shafer theory, (3) fuzzy set theory, (4) MYCIN Model, (5) Cohen's system of inductive probabilities, and (6) a class of non- monotonic reasoning methods. Each method is presented and discussed in terms of theoretical content, a detailed numerical example, and a list of strengths and limitations. Purposely, the same numerical example is addressed by each method such that we are able to highlight the assumptions, and computational requirements that are specific to each method in a consistent manner.