Interval-based easoning in Medical Diagnosis
Kok Meng Yew, Kohout · 1997
1.0 Introduction In this paper, we look at inference based on subset containment, supported by the fuzzy power set theory (2) and mathematical relations. Interval-valued inference structures used here were developed in (2,5,7) and were based on sub triangle products ( 11 and their modifications. The interval-valued inference studied here is based on the checklist paradigm (4). A simulator was built using methodology (3) to evaluate these fuzzy relational inference structures in a complete medical domain which was based on CLINAID (6). The simulation used fuzzy input data, computation with and without the material paradox of implication in body systems identification. We used inference bands for determining acceptance and rejection (7) and performance metrics in (8) for measuring the degree of acceptance and rejection. Together, they provide the ranking of the inference structures. The performance of the inference structures based on these metrics were reported. 2.0 Inference Structures In this paper, we are interested in the SUB inference structures (7) using fuzzy implication measure ml (l) of the checklist paradigm. Given that R and S are binary relations, the following inference structures were evaluated :