Causal and plausible reasoning in expert systems

摘要: This research is directed at the development of a better understanding roles causal and plausible reasoning in management uncertainty expert systems. In earlier studies, these modes were considered as separate issues, with dissociation aspects from an assessment degree likelihood. present study, question are analyzed unified point view, yielding new type representation called structured rules that reflect both inference causation strength. Causation endorsed through set relationships known roles, including "sufficient," "associational," "supportive," "weak" "strong necessary," "contrary," "exceptional," whereas strength measured by "conditional basic probability assignment" associated conclusion, much Bayesian conditional addresses uncertain rules. Each role describes qualitatively special form inference. addition, different when combined for same rule, produce body coherent knowledge represented locally. this way, normal associational relationship can be augmented several supportive or exception conditions. comparison to conventional rules, such localized structure facilitates more focused acquisition simplifies task rule interpretation conclusion has modified later process. The Dempster-Shafer theory selected basis because its generality ability deal incomplete information. The original extended support chains combine conclusions multiple prior information available. Practical solutions problem imprecise concepts also developed. With extensions, measures beliefs expressed context individual roles; during reasoning, ignorance then propagated inferences. work generalizes results on evidential reasoning. A prototype knowledge-based system venture investment evaluation been implemented KEE. Its purpose demonstrate under uncertainly based experiment frames slots.